Image and Graphics 9th International Conference ICIG 2017 Shanghai China September 13 15 2017 Revised Selected Papers Part III 1st Edition Yao Zhao

Image and Graphics 9th International
Conference ICIG 2017 Shanghai China

September 13 15 2017 Revised Selected
Papers Part III 1st Edition Yao Zhao
Visit to download the full and correct content document:
https://textbookfull.com/product/image-and-graphics-9th-international-conference-icig-
2017-shanghai-china-september-13-15-2017-revised-selected-papers-part-iii-1st-editi
on-yao-zhao/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...
Image and Graphics 10th International Conference ICIG

2019 Beijing China August 23 25 2019 Proceedings Part
III Yao Zhao
https://textbookfull.com/product/image-and-graphics-10th-
international-conference-icig-2019-beijing-china-
august-23-25-2019-proceedings-part-iii-yao-zhao/

2015 Tianjin China August 13 16 2015 Proceedings Part
III 1st Edition Yu-Jin Zhang (Eds.)
international-conference-icig-2015-tianjin-china-
august-13-16-2015-proceedings-part-iii-1st-edition-yu-jin-zhang-
eds/
Advances in Image and Graphics Technologies 12th

Chinese conference IGTA 2017 Beijing China June 30 July
1 2017 Revised Selected Papers 1st Edition Yongtian
Wang
https://textbookfull.com/product/advances-in-image-and-graphics-
technologies-12th-chinese-conference-igta-2017-beijing-china-
june-30-july-1-2017-revised-selected-papers-1st-edition-yongtian-
wang/
Smart Card Research and Advanced Applications 16th

International Conference CARDIS 2017 Lugano Switzerland
November 13 15 2017 Revised Selected Papers 1st Edition
Thomas Eisenbarth
https://textbookfull.com/product/smart-card-research-and-
advanced-applications-16th-international-conference-
cardis-2017-lugano-switzerland-november-13-15-2017-revised-
Internet Multimedia Computing and Service 9th
International Conference ICIMCS 2017 Qingdao China
August 23 25 2017 Revised Selected Papers 1st Edition
Benoit Huet
https://textbookfull.com/product/internet-multimedia-computing-
and-service-9th-international-conference-icimcs-2017-qingdao-
china-august-23-25-2017-revised-selected-papers-1st-edition-
benoit-huet/
Computational Intelligence and Intelligent Systems 9th

International Symposium ISICA 2017 Guangzhou China
November 18 19 2017 Revised Selected Papers Part II
Kangshun Li
https://textbookfull.com/product/computational-intelligence-and-
intelligent-systems-9th-international-symposium-
isica-2017-guangzhou-china-november-18-19-2017-revised-selected-
papers-part-ii-kangshun-li/

2015 Tianjin China August 13 16 2015 Proceedings Part I
1st Edition Yu-Jin Zhang (Eds.)
international-conference-icig-2015-tianjin-china-
august-13-16-2015-proceedings-part-i-1st-edition-yu-jin-zhang-
eds/
Image Analysis and Processing ICIAP 2017 19th

International Conference Catania Italy September 11 15
2017 Proceedings Part II 1st Edition Sebastiano
Battiato
https://textbookfull.com/product/image-analysis-and-processing-
iciap-2017-19th-international-conference-catania-italy-
september-11-15-2017-proceedings-part-ii-1st-edition-sebastiano-
battiato/
Image Analysis and Processing ICIAP 2017 19th

International Conference Catania Italy September 11 15
2017 Proceedings Part I 1st Edition Sebastiano Battiato
https://textbookfull.com/product/image-analysis-and-processing-
iciap-2017-19th-international-conference-catania-italy-
september-11-15-2017-proceedings-part-i-1st-edition-sebastiano-
Yao Zhao · Xiangwei Kong
David Taubman (Eds.)
LNCS 10668
Image
and Graphics
9th International Conference, ICIG 2017
Shanghai, China, September 13–15, 2017
Revised Selected Papers, Part III
123
Lecture Notes in Computer Science 10668
Commenced Publication in 1973
Founding and Former Series Editors:
Gerhard Goos, Juris Hartmanis, and Jan van Leeuwen
Editorial Board
David Hutchison
Lancaster University, Lancaster, UK
Takeo Kanade
Carnegie Mellon University, Pittsburgh, PA, USA
Josef Kittler
University of Surrey, Guildford, UK
Jon M. Kleinberg
Cornell University, Ithaca, NY, USA
Friedemann Mattern
ETH Zurich, Zurich, Switzerland
John C. Mitchell
Stanford University, Stanford, CA, USA
Moni Naor
Weizmann Institute of Science, Rehovot, Israel
C. Pandu Rangan
Indian Institute of Technology, Madras, India
Bernhard Steffen
TU Dortmund University, Dortmund, Germany
Demetri Terzopoulos
University of California, Los Angeles, CA, USA
Doug Tygar
University of California, Berkeley, CA, USA
Gerhard Weikum
Max Planck Institute for Informatics, Saarbrücken, Germany
More information about this series at http://www.springer.com/series/7412
Yao Zhao Xiangwei Kong
•
David Taubman (Eds.)
Image
and Graphics
9th International Conference, ICIG 2017
Shanghai, China, September 13–15, 2017
Revised Selected Papers, Part III
123
Editors
Yao Zhao David Taubman
Beijing Jiaotong University UNSW
Beijing Sydney, NSW
China Australia
Xiangwei Kong
Dalian University of Technology
Dalian
China
ISSN 0302-9743 ISSN 1611-3349 (electronic)

Lecture Notes in Computer Science
ISBN 978-3-319-71597-1 ISBN 978-3-319-71598-8 (eBook)
https://doi.org/10.1007/978-3-319-71598-8
Library of Congress Control Number: 2017960877
LNCS Sublibrary: SL6 – Image Processing, Computer Vision, Pattern Recognition, and Graphics
© Springer International Publishing AG 2017

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the
material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now
known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book are
believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors
give a warranty, express or implied, with respect to the material contained herein or for any errors or
omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.
Printed on acid-free paper
This Springer imprint is published by Springer Nature

The registered company is Springer International Publishing AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface
These are the proceedings of the 8th International Conference on Image and Graphics
(ICIG 2017), held in Shanghai, China, during September 13–15, 2017.
The China Society of Image and Graphics (CSIG) has hosted this series of ICIG
conferences since 2000. ICIG is the biennial conference organized by the China
Society of Image and Graphics (CSIG), focusing on innovative technologies of image,
video, and graphics processing and fostering innovation, entrepreneurship, and net-
working. This time, Shanghai Jiaotong University was the organizer, and the Nanjing
Technology University and Zong Mu Technology Ltd. Company were the
co-organizers. Details about the past eight conferences, as well as the current one, are
as follows:
Conference Place Date Submitted Proceeding

First (ICIG 2000) Tianjin, China August 16–18 220 156
Second (ICIG 2002) Hefei, China August 15–18 280 166
Third (ICIG 2004) Hong Kong, China December 17–19 460 140
4th (ICIG 2007) Chengdu, China August 22–24 525 184
5th (ICIG 2009) Xi’an, China September 20–23 362 179
6th (ICIG 2011) Hefei, China August 12–15 329 183
7th (ICIG 2013) Qingdao, China July 26–28 346 181
8th (ICIG 2015) Tianjin, China August 13–16 345 170
9th (ICIG 2017) Shanghai, China September 13–15 370 172
This time, the proceedings are published by Springer in the LNCS series. The titles,
abstracts, and biographies of the three invited speakers of plenary talks are presented
first. At ICIG 2017, 370 submissions were received, and 160 papers were accepted. To
ease in the search of a required paper in these proceedings, the 160 regular papers have
been arranged in alphabetical order according to their titles. Another 12 papers forming
a special topic are included at the end.
Our sincere thanks go to all the contributors (around 200), who came from around
the world to present their advanced works at this event. Special thanks go to the
members of Technical Program Committee, who carefully reviewed every single
submission and made their valuable comments for improving the accepted papers.
The proceedings could not have been produced without the invaluable efforts of the
publication chairs, the web chairs, and a number of active members of CSIG.
September 2017 Yao Zhao

Xiangwei Kong
David Taubman
Organization
Honorary Chairs
Guanhua Xu MOST, China
Yuan F. Zheng Ohio State University, USA
General Chairs
Tieniu Tan Chinese Academy of Sciences, China
Hongkai Xiong Shanghai Jiaotong University, China
Zixiang Xiong Texas A&M University, USA
Organizing Committee Chairs

Weiyao Lin Shanghai Jiaotong University, China
Huimin Ma Tsinghua University, China
Bo Yan Fudan University, China
Technical Program Chairs

David Taubman UNSW, Australia
Yao Zhao Beijing Jiaotong University, China
Finance Chairs
Zhihua Chen ECUST, China
Zhenwei Shi Beihang University, China
Special Session Chairs

Jian Cheng Chinese Academy of Sciences, China
Zhihai He University of Missouri, USA
Z. Jane Wang University of British Columbia, Canada
Award Chairs
Xin Li West Virginia University, USA
Shiqiang Yang Tsinghua University, China
VIII Organization
Publicity Chairs
Mingming Cheng Nankai University, China
Moncef Gabbouj TUT, Finland
Exhibits Chairs
Zhijun Fang Shanghai University of Engineering Science, China
Yan Lv Microsoft Research, China
Publication Chairs
Xiangwei Kong Dalian University of Technology, China
Jun Yan Journal of Image and Graphics, China
International Liaisons
Xiaoqian Jiang UCSD, USA
Huifang Sun MERL, USA
Local Chairs
Wenrui Dai UCSD, USA
Junni Zou Shanghai Jiaotong University, China
Registration Chair
Chen Ye Shanghai Jiaotong University, China
Webmasters
Chenglin Li EPFL, Switzerland
Yangmei Shen Shanghai Jiaotong University, China
Technical Program Committee

Ping An Shanghai University, China
Ru An Hohai University, China
Xiao Bai Beijing University of Aeronautics and Astronautics, China
Lianfa Bai Nanjing University of Science and Technology, China
Xiang Bai Huazhong University of Science and Technology, China
Chongke Bi Tianjin University, China
Hai Bian Hangzhou Dica3d Technology Co., Ltd., China
Xiaochun Cao Institute of Information Engineering,
Chinese Academy of Sciences, China
Organization IX
Yan-Pei Cao Tsinghua University, China

Chong Cao Tsinghua University, China
Qi Chen Hainan University, China
Kang Chen Tsinghua University, China
Mingkai Chen Nanjing University of Posts and Telecommunications, China
Mingming Cheng Nankai University, China
Yue Dong MSRA, China
Zhijun Fang Shanghai University of Engineering Science, China
Qianjin Feng Southern Medical University, China
Xiaoyi Feng Northwestern Polytechnical University, China
Dongmei Fu University of Science and Technology Beijing, China
Junying Gan Wuyi University, China
Lin Gao ICT, CAS, China
Yue Gao Tsinghua University, China
Xinbo Gao Xidian University, China
Zexun Geng Information Engineering University, China
Guanghua Gu Yanshan University, China
Lin Gu National Institute of Informatics, Japan
Yanwen Guo Nanjing University, China
Hu Han Nanyang Technological University, Singapore
Xiaowei He Northwest University, China
Qiming Hou Zhejiang University, China
Dong Hu Nanjing University of Posts and Telecommunications, China
Hua Huang Beijing Institute of Technology, China
Haozhi Huang Tsinghua University, China
Yongfeng Huang Tsinghua University, China
Rongrong Ji Xiamen University, China
Yunde Jia Beijing Institute of Technology, China
Sen Jia Shenzhen University, China
Xiuping Jia University of New South Wales, USA
Zhiguo Jiang Beijing University of Aeronautics and Astronautics, China
Zhaohui Jiang Central South University, China
Xiaoqian Jiang University of California, San Diego, USA
Lianwen Jin South China University of Technology, China
Bin Kong Institute of Intelligent Machines,
Xiangwei Kong Dalian University of Technology, China
Dengfeng Kuang Nankai University, China
Jianhuang Lai Sun Yat-Sen University, China
Congyan Lang Beijing Jiaotong University, China
Changhua Li Xi’an University of Architecture and Technology, China
Chenglin Li Swiss Federal Institute of Technology in Lausanne,
Switzerland
Hua Li Institute of Computing Technology, Chinese Academy
of Sciences, China
Jiming Li Zhejiang Police College, China
X Organization
Qi Li Peking University, China

Shutao Li Hunan University, China
Xi Li Zhejiang University, China
Jie Liang China Aerodynamics Research and Development Center,
China
Pin Liao Nanchang University, China
Chunyu Lin Beijing Jiaotong University, China
Xiaojing Liu Qinghai University, China
Changhong Liu Jiangxi Normal University, China
Bin Liu University of Science and Technology of China, China
Bin Liu Tsinghua University, China
Chenglin Liu Institute of Automation, Chinese Academy of Sciences, China
Wenyu Liu Huazhong University of Science and Technology, China
Yue Liu Beijing Institute of Technology, China
Qingshan Liu Nanjing University of Information Science and Technology,
China
Hongbing Lu Fourth Military Medical University, China
Hanqing Lu Institute of Automation, Chinese Academy of Sciences, China
Jiwen Lu Tsinghua University, China
Jianhua Ma Southern Medical University, China
Huimin Ma Tsinghua University, China
Weidong Min Nanchang University, China
Xuanqin Mou Xi’an Jiaotong University, China
Taijiang Mu Tsinghua University, China
Feiping Nie Northwestern Polytechnical University, China
Yongwei Nie South China University of Technology, China
Zhigeng Pan Hangzhou Normal University, China
Yanwei Pang Tianjin University, China
Yuxin Peng Peking University, China
Yuntao Qian Zhejiang University, China
Bo Ren Nankai University, China
Jun Sang Chongqing University, China
Nong Sang Huazhong University of Science and Technology, China
Yangmei Shen Shanghai Jiaotong University, China
Yuying Shi North China Electric Power University, China
Huifang Sun Mitsubishi Electric Research Laboratories, USA
Jiande Sun Shandong University, China
Linmi Tao Tsinghua University, China
Lei Tong Beijing University of Technology, China
Yunhai Wang Shandong University, China
Qi Wang Northwestern Polytechnical University, China
Cheng Wang Xiamen University, China
Meng Wang Hefei University of Technology, China
Hanzi Wang Xiamen University, China
Peizhen Wang Anhui University of Technology, China
Tianjiang Wang Huazhong University of Science and Technology, China
Organization XI
Bin Wang Tsinghua University, China

Lili Wang Beihang University, China
Shigang Wang Jilin University, China
Miao Wang Tsinghua University, China
Yunhong Wang Beijing University of Aeronautics and Astronautics, China
Chunhong Wu University of Science and Technology Beijing, China
Hongzhi Wu Zhejiang University, China
Xiaojun Wu Jiangnan University, China
Fei Wu Zhejiang University, China
Zhongke Wu Beijing Normal University, China
Dingyuan Xia Wuhan University of Technology, China
Hongkai Xiong Shanghai Jiaotong University, China
Mingliang Xu Zhengzhou University, China
Chunxu Xu Tsinghua University, China
Kun Xu Tsinghua University, China
Zengpu Xu Tianjin University of Science and Technology, China
Jianru Xue Xi’an Jiaotong University, China
Xiangyang Xue Fudan University, China
Bo Yan Fudan University, China
Ling-Qi Yan UC Berkeley, USA
Xiao Yan Tsinghua University, China
Jingwen Yan Shantou University, China
Jun Yan Institute of Remote Sensing and Digital Earth,
Jinfeng Yang Civil Aviation University of China, China
Sheng Yang Tsinghua University, China
Yongliang Yang Bath University, UK
Shiqiang Yang Tsinghua University, China
Tao Yang Tsinghua University, China
Hongxun Yao Harbin Institute of Technology, China
Yong Yin Dalian Maritime University, China
Shiqi Yu Shenzhen University, China
Nenghai Yu University of Science and Technology of China, China
Yinwei Zhan Guangdong University of Technology, China
Aiqing Zhang Anhui Normal University, China
Wei Zhang Shandong University, China
Daoqiang Zhang Nanjing University of Aeronautics and Astronautics, China
Jiawan Zhang Tianjin University, China
Lei Zhang Beijing Institute of Technology, China
Song-Hai Zhang Tsinghua University, China
Shiliang Zhang Peking University, China
Xinpeng Zhang Shanghai University, China
Yanci Zhang Sichuan University, China
Yongfei Zhang Beijing University of Aeronautics and Astronautics, China
Fang-Lue Zhang Victoria University of Wellington, New Zealand
Guofeng Zhang Zhejiang University, China
XII Organization
Qiang Zhang Dalian University, China

Yun Zhang Zhejiang University of Media and Communications, China
Liangpei Zhang Wuhan University, China
Shengchuan Zhang Xiamen University, China
Xiaopeng Zhang Shanghai Jiaotong University, China
Sicheng Zhao Tsinghua University, China
Yao Zhao Beijing Jiaotong University, China
Jieyu Zhao Ningbo University, China
Chunhui Zhao Harbin Engineering University, China
Ying Zhao Central South University, China
Wei-Shi Zheng Sun Yat-Sen University, China
Ping Zhong National University of Defense Technology, China
Quan Zhou China Academy of Space Technology, Xi’an, China
Jun Zhou Griffith University, Australia
Liang Zhou Nanjing University of Posts and Telecommunications, China
Linna Zhou University of International Relations, China
Tao Zhou Ningxia Medical University, China
Wengang Zhou University of Science and Technology of China, China
Zhe Zhu Duke University, USA
Wang-Jiang Zhu Tsinghua University, China
Yonggui Zhu Communication University of China, China
Contents – Part III
Computational Imaging
A Double Recursion Algorithm to Image Restoration

from Random Limited Frequency Data . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Xiaoman Liu and Jijun Liu
RGB-D Saliency Detection with Multi-feature-fused Optimization . . . . . . . . . 15

Tianyi Zhang, Zhong Yang, and Jiarong Song
Research on Color Image Segmentation . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

Jingwen Yan, Xiaopeng Chen, Tingting Xie,
and Huimin Zhao
Depth-Based Focus Stacking with Labeled-Laplacian Propagation . . . . . . . . . 36

Wentao Li, Guijin Wang, Xuanwu Yin, Xiaowei Hu,
and Huazhong Yang
A Novel Method for Measuring Shield Tunnel Cross Sections . . . . . . . . . . . 47

Ya-dong Xue, Sen Zhang, and Zhe-ting Qi
Image Noise Estimation Based on Principal Component

Analysis and Variance-Stabilizing Transformation . . . . . . . . . . . . . . . . . . . . 58
Ling Ding, Huying Zhang, Bijun Li, Jinsheng Xiao,
and Jian Zhou
Efficient High Dynamic Range Video Using Multi-exposure CNN Flow . . . . 70

Yuchen Guo, Zhifeng Xie, Wenjun Zhang,
and Lizhuang Ma
Depth Image Acquisition Method in Virtual Interaction

of VR Yacht Simulator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
Qin Zhang and Yong Yin
An Image Segmentation Method Based on Asynchronous

Multiscale Similarity Measure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
Min Li, Zhongwai Xu, Hongwen Xie,
and Yuhang Xing
Lagrange Detector in Image Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . 102

Feilong Ma, Linmi Tao, and Wu Xia
New Tikhonov Regularization for Blind Image Restoration . . . . . . . . . . . . . 113

Yuying Shi, Qiao Liu, and Yonggui Zhu
XIV Contents – Part III
Real-Time Multi-camera Video Stitching

Based on Improved Optimal Stitch Line and Multi-resolution Fusion. . . . . . . 124
Dong-Bin Xu, He-Meng Tao, Jing Yu, and Chuang-Bai Xiao
Image Quality Assessment of Enriched Tonal Levels Images . . . . . . . . . . . . 134

Jie Zhao, Wei Wen, and Siamak Khatibi
Computer Graphics and Visualization
A Variational Model to Extract Texture from Noisy Image

Data with Local Variance Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
Tao Zhang and Qiuli Gao
Joint Visualization of UKF Tractography Data . . . . . . . . . . . . . . . . . . . . . . 158

Wen Zhao, Wenyao Zhang, Na Wang,
and Pin Liao
Semantic Segmentation Based Automatic Two-Tone Portrait Synthesis . . . . . 170

Zhuoqi Ma, Nannan Wang, Xinbo Gao, and Jie Li
Parameters Sharing Multi-items Non-parametric

Factor Microfacet Model for Isotropic and Anisotropic BRDFs . . . . . . . . . . . 182
Junkai Peng, Changwen Zheng,
and Pin Lv
SRG and RMSE-Based Automated Segmentation for Volume Data . . . . . . . . 194

Wang Li, Xiaoan Tang, and Junda Zhang
Shape Recovery of Endoscopic Videos by Shape from Shading

Using Mesh Regularization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
Zhihang Ren, Tong He, Lingbing Peng, Shuaicheng Liu,
Shuyuan Zhu, and Bing Zeng
Lazy Recoloring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214

Guanlei Xu, Xiaotong Wang, Xiaogang Xu, and Lijia Zhou
Similar Trademark Image Retrieval Integrating LBP

and Convolutional Neural Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
Tian Lan, Xiaoyi Feng, Zhaoqiang Xia, Shijie Pan,
and Jinye Peng
Adaptive Learning Compressive Tracking Based on Kalman Filter . . . . . . . . 243

Xingyu Zhou, Dongmei Fu, Yanan Shi, and Chunhong Wu
Online High-Accurate Calibration of RGB+3D-LiDAR

for Autonomous Driving . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254
Tao Li, Jianwu Fang, Yang Zhong, Di Wang, and Jianru Xue
Contents – Part III XV
Run-Based Connected Components Labeling Using Double-Row Scan . . . . . 264

Dongdong Ma, Shaojun Liu, and Qingmin Liao
A 3D Tube-Object Centerline Extraction Algorithm

Based on Steady Fluid Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275
Dongjin Huang, Ruobin Gong, Hejuan Li, Wen Tang,
and Youdong Ding
Moving Objects Detection in Video Sequences Captured

by a PTZ Camera . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287
Li Lin, Bin Wang, Fen Wu, and Fengyin Cao
Fast Grid-Based Fluid Dynamics Simulation

with Conservation of Momentum and Kinetic Energy on GPU . . . . . . . . . . . 299
Ka-Hou Chan and Sio-Kei Im
Adaptive Density Optimization of Lattice Structures Sustaining

the External Multi-load . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311
Li Shi, Changdong Zhang, Tingting Liu, Wenhe Liao,
and Xiuyi Jia
Hyperspectral Image Processing
Hyperspectral Image Classification Based on Deep Forest

and Spectral-Spatial Cooperative Feature . . . . . . . . . . . . . . . . . . . . . . . . . . 325
Mingyang Li, Ning Zhang, Bin Pan, Shaobiao Xie, Xi Wu,
and Zhenwei Shi
Hyperspectral Image Classification Using Multi Vote Strategy

on Convolutional Neural Network and Sparse Representation
Joint Feature. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337
Daoming Ye, Rong Zhang, and Dixiu Xue
Efficient Deep Belief Network Based Hyperspectral Image Classification . . . . 347

Atif Mughees and Linmi Tao
Classification of Hyperspectral Imagery Based on Dictionary Learning

and Extended Multi-attribute Profiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 358
Qishuo Gao, Samsung Lim, and Xiuping Jia
Deep Residual Convolutional Neural Network for Hyperspectral

Image Super-Resolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 370
Chen Wang, Yun Liu, Xiao Bai, Wenzhong Tang, Peng Lei,
and Jun Zhou
XVI Contents – Part III
Multi-view and Stereoscopic Processing
Stereoscopic Digital Camouflage Pattern Generation Algorithm

Based on Color Image Segmentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 383
Qin Lei, Wei-dong Xu, Jiang-hua Hu, and Chun-yu Xu
Uncertain Region Identification for Stereoscopic Foreground Cutout . . . . . . . 390

Taotao Yang, Shuaicheng Liu, Chao Sun, Zhengning Wang,
and Bing Zeng
Map-Build Algorithm Based on the Relative Location of Feature Points . . . . 400

Cheng Zhao, Fuqiang Zhao, and Bin Kong
Sparse Acquisition Integral Imaging System . . . . . . . . . . . . . . . . . . . . . . . . 412

Bowen Jia, Shigang Wang, Wei Wu, Tianshu Li,
and Lizhong Zhang
Marker-Less 3D Human Motion Capture in Real-Time

Using Particle Swarm Optimization with GPU-Accelerated
Fitness Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 423
Bogdan Kwolek and Boguslaw Rymut
Warping and Blending Enhancement for 3D View Synthesis

Based on Grid Deformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 436
Ningning Hu, Yao Zhao, and Huihui Bai
A Vehicle-Mounted Multi-camera 3D Panoramic Imaging Algorithm

Based on Ship-Shaped Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 445
Xin Wang, Chunyu Lin, Yi Gao, Yaru Li, Shikui Wei,
and Yao Zhao
A Quality Evaluation Scheme to 3D Printing Objects

Using Stereovision Measurement. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 458
Li-fang Wu, Xiao-hua Guo, Li-dong Zhao,
and Meng Jian
Representation, Analysis and Applications of Large-Scale

3D Multimedia Data
Secure Image Denoising over Two Clouds . . . . . . . . . . . . . . . . . . . . . . . . . 471

Xianjun Hu, Weiming Zhang, Honggang Hu,
and Nenghai Yu
Aesthetic Quality Assessment of Photos with Faces. . . . . . . . . . . . . . . . . . . 483

Weining Wang, Jiexiong Huang, Xiangmin Xu,
Quanzeng You, and Jiebo Luo
Contents – Part III XVII
Sensitive Information Detection on Cyber-Space . . . . . . . . . . . . . . . . . . . . . 496

Mingbao Lin, Xianming Lin, Yunhang Shen,
and Rongrong Ji
Optimization Algorithm Toward Deep Features

Based Camera Pose Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 505
Han Chen, Feng Guo, Ying Lin,
and Rongrong Ji
Security
A Robust 3D Video Watermarking Scheme

Based on Multi-modal Visual Redundancy . . . . . . . . . . . . . . . . . . . . . . . . . 517
Congxin Cheng, Wei Ma, Yuchen Yang,
Shiyang Zhang, and Mana Zheng
Partial Secret Image Sharing for (n, n) Threshold

Based on Image Inpainting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 527
Xuehu Yan, Yuliang Lu, Lintao Liu, Shen Wang,
Song Wan, Wanmeng Ding, and Hanlin Liu
A New SMVQ-Based Reversible Data Hiding Scheme

Using State-Codebook Sorting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 539
Juan-ni Liu, Quan Zhou, Yan-lang Hu,
and Jia-yuan Wei
An Efficient Privacy-Preserving Classification Method

with Condensed Information. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 551
Xinning Li and Zhiping Zhou
Cross-Class and Inter-class Alignment Based Camera Source

Identification for Re-compression Images . . . . . . . . . . . . . . . . . . . . . . . . . . 563
Guowen Zhang, Bo Wang, and Yabin Li
JPEG Photo Privacy-Preserving Algorithm Based on Sparse

Representation and Data Hiding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 575
Wenjie Li, Rongrong Ni, and Yao Zhao
Surveillance and Remote Sensing
An Application Independent Logic Framework

for Human Activity Recognition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 589
Wengang Feng, Yanhui Xiao, Huawei Tian, Yunqi Tang,
and Jianwei Ding
XVIII Contents – Part III
An Altitude Based Landslide and Debris Flow Detection Method

for a Single Mountain Remote Sensing Image . . . . . . . . . . . . . . . . . . . . . . 601
Tingting Sheng and Qiang Chen
Improved Fully Convolutional Network for the Detection

of Built-up Areas in High Resolution SAR Images . . . . . . . . . . . . . . . . . . . 611
Ding-Li Gao, Rong Zhang, and Di-Xiu Xue
SAR Image Registration with Optimized Feature Descriptor

and Reliable Feature Matching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 621
Yanzhao Wang, Juan Su, Bichao Zhan, Bing Li, and Wei Wu
An Improved Feature Selection Method for Target Discrimination

in SAR Images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 633
Yanyan Li and Aihua Cai
The Detection of Built-up Areas in High-Resolution SAR Images

Based on Deep Neural Networks. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 646
Yunfei Wu, Rong Zhang, and Yue Li
SAR Automatic Target Recognition Based on Deep Convolutional

Neural Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 656
Ying Xu, Kaipin Liu, Zilu Ying, Lijuan Shang, Jian Liu,
Yikui Zhai, Vincenzo Piuri, and Fabio Scotti
Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 669

Computational Imaging
A Double Recursion Algorithm to Image
Restoration from Random Limited
Frequency Data
Xiaoman Liu and Jijun Liu(B)
School of Mathematics, Southeast University, Nanjing 211189,

People’s Republic of China
jjliu@seu.edu.cn
Abstract. One of the main tasks in image restoration is to catch the

picture characteristics such as interfaces and textures from incomplete
noisy frequency data. For the cost functional with data matching term in
frequency domain and the total variation together with Frobenius norm
penalty terms in spatial domain, the properties of the minimizer of cost
functional and the error estimates on the regularizing solution are estab-
lished. Then we propose an algorithm with double recursion to restore
piecewise smooth image. The Bregman iteration with lagged diffusivity
fixed point method is used to solve the corresponding nonlinear Euler-
Lagrange equation. By implementing recursion algorithms a few times,
the satisfactory reconstructions can be obtained using random band sam-
pling data. Numerical implementations demonstrate the validity of our
proposed algorithm with good edge-preservations.
Keywords: Image restoration · Random sampling · Total variation

Bregman iteration · Adjoint conjugate gradient method
1 Introduction
A lot of engineering backgrounds such as objects detections for military pur-

poses and magnetic resonance imaging (MRI) for medical applications, lead to
the studies on image restorations. The main tasks of image restorations aim
to the recovery of an image from given noisy measurement data. However, in
most engineering configurations, the specified measurement data are limited fre-
quency data instead of the full data in spatial domain. Therefore the image
restorations are essentially of the nature of ill-posedness [1]. Mathematically,
the image restorations require the stable approximations to the target image
from insufficient noisy measurement data, which are also one of the most impor-
tant research areas of applied mathematics by dealing with the ill-posedness of
the image restorations.
In recent decades, various image restoration models and mathematical tech-
niques have been developed due to the great importance of image restoration
c Springer International Publishing AG 2017
Y. Zhao et al. (Eds.): ICIG 2017, Part III, LNCS 10668, pp. 3–14, 2017.
https://doi.org/10.1007/978-3-319-71598-8_1
4 X. Liu and J. Liu
problems, such as level set methods, wavelet-based frameworks, nonlinear PDEs

models and optimization schemes [2–4]. In all these studies, the basic ideas are
the reconstructions of an image by some denoising process, while the key infor-
mation about the image should be kept in terms of the regularization techniques.
The mathematical framework dealing with ill-posed problems is the regular-
izing scheme with some appropriate penalty terms incorporating into the cost
functional. The key issue for this scheme is that the suitable weight called the
regularizing parameters between the data matching term and the penalty terms
should be specified artificially to keep the balance between the data matching
and smoothness of the sought solution. In the cases that the exact solution to be
sought is smooth, the penalty terms for the solution can be measured by some
standard norms such as L2 or H 2 norm for which the choice strategy for the
regularizing parameters have been studied thoroughly [1,4]. However, for non-
smooth exact solution such as those in the image restorations with the sharp
jump of the grey level function of the image, instead of the standard differential
norms, the other non-differential penalty term such as total variation (TV) or
L0 -norm sparsity penalty term should be applied [2,3].
Motivated by the above engineering backgrounds, we consider an image
recovery problem using the incomplete noisy frequency data by minimizing a
cost functional with penalty terms in Sect. 2. Based on the derivatives expres-
sions of the cost functional, an iterative scheme with outer and inner recursions
are proposed in Sect. 3 to solve the minimizing problem. Finally, some numerical
experiments are presented to show the validity of the proposed scheme in Sect. 4.
2 Optimization Modeling with Error Analysis

As the standard configurations in signal processing, we define the Fourier Trans-
form matrix F ∈ RN ×N with the components
2π
Fi,j = e−i N ij , i, j = 1, · · · , N. (2.1)
It is well-known that the matrix F is an unitary matrix [10] satisfying F ∗ F =

I, where the superscript ∗ denotes the conjugate transpose of an matrix, I is
identity matrix.
In most computer vision problems, the two-dimensional image f :=
(fm,n )(m, n = 1, · · · , N ) can be represented as a vector f . Here we introduce
some symbols for this representation as follows:
operator
2
vect : RN ×N → RN ×1 : vect[f ] := (f1 , f2 , · · · , fN 2 )T = f , where
2
the N -elements are generated by re-ordering the N column vectors of f
sequently.
operator array: the inverse of the vect, i.e., array[f ] = f .
Two-dimensional discrete Fourier transform (DFT) matrix F:
f̂ := vect[F T f F ] = (F ⊗ F )f := Ff ,
where ⊗ is the tensor product of two matrices.

A Double Recursion Algorithm to Image Restoration 5
By partial frequency data of fˆ, we mean that only parts of the elements of
fˆ are sampled, with other elements considered as 0. Denoted by P the N × N
matrix generating from the identity matrix I by setting its N –M rows as null
vectors, i.e., P = diag(p11 , p22 , · · · , pN N ) with pii being 1 or 0. Then P fˆ means
that we only take M (≤ N ) rows of fˆ as our partial data. To the vector form f ,
the sampling matrix should be modified as an N 2 × N 2 matrix. Then we have
the following notations with tensor product:
vect[P fˆ] = (I ⊗ P )vect[fˆ] = (I ⊗ P )f̂ := Pf̂ . (2.2)
Obviously, the matrix P is the sampling matrix in algorithm domain which is
chosen before reconstruction.
Generally, the frequency data of an image f are obtained by some scanning
process, which are of unavoidable error, i.e., our inversion data for image recovery
are in fact P ĝ δ with noisy data ĝ δ of fˆ satisfying
P fˆ − P ĝ δ F ≤ fˆ − ĝ δ F ≤ δ, (2.3)
where · F is the Frobenius norm for an N × N matrix, corresponding to the
2-norm of an N 2 -dimensional vector after stacking the elements of an N × N
matrix into an N 2 -dimensional vector. Hence the data matching term can be
written as
P fˆ − P ĝ δ 2F = P(Ff ) − Pĝδ 22 . (2.4)
There are various sampling matrices, like those in radial lines sampling
method [12], band sampling method [6] which is much efficient in numerical
experiments. In this paper, we apply the random band sampling process which
samples some rows randomly. Denoted by cenR the central ratio, i.e., to the fre-
quency image there are only cenR × N rows in the central parts of the frequency
image (the centre is between No. N/2−1 and N /2 rows) in Cartesian coordinate
system or in the natural domain coordinate which is shown in Fig. 1(b), and the
others sampling rows are the random ones. In the algorithm domain shown in
Fig. 1(a), the sampling rows are distributed on the four corners, i.e., it is the
same distribution as the mask as usual.
Fig. 1. (a) Algorithm domain coordinate; (b) Natural domain coordinate.
Recall that the most important issue for image restorations is the edge-
preservation property of the image, which means that we are essentially inter-
ested in the efficient reconstruction for a piecewise constant image with the
6 X. Liu and J. Liu
main interests in detecting image edges. Since the total variation (TV) of a two-
dimensional function can describe the function jumps efficiently, we are then led
to the following constraint optimization problem

min |f |T V : P F T f F − P ĝ δ 2F ≤ δ , (2.5)
f ∈RN ×N
where P ĝ δ is the incomplete noisy frequency data, and TV penalty term |f |T V

for f is defined in a standard way [6,12]. Since |f |T V is not differentiable at
f = Θ (zero matrix), we approximate |f |T V by

N
2 2
|f |T V,β = (∇xm,n
1
f ) + (∇xm,n
2
f) + β (2.6)
m,n=1

for small constant β > 0, where ∇m,n f := ∇xm,n
1
f, ∇xm,n
2
f with two components

fm+1,n − fm,n , if m < N, x2 fm,n+1 − fm,n , if n < N,
∇xm,n
1
f = ∇m,n f =
f1,n − fm,n , if m = N, fm,1 − fm,n , if n = N
for m, n = 1, · · · , N due to the periodic boundary condition on f .

However, the constraint optimization problem (2.5) in the case of P = I has
no restrictions on the size of f , notice that P F T XF = Θ may have nonzero
solution X arbitrarily large for singular matrix P . To exclude this uncertainty,
our image recovery problem is finally reformulated as the following unconstraint
problem
∗
f := arg min Jβ (f ),
f (2.7)
Jβ (f ) := 12 P F T f F − P ĝ δ 2F + α1 f 2F + α2 |f |T V,β .
where α1 , α2 > 0 are regularizing parameters.

The theorems below illustrate the existence of the minimizer and establish
the choice strategies for the regularizing parameters α1 , α2 .
Theorem 1. For α1 > 0, α2 , β ≥ 0, there exists a local minimizer to the opti-

mization problem (2.7).
Proof. Since Jβ (f ) ≥ 0 for f ∈ RN ×N , there exists a constant J ∗ ≥ 0 such that

J ∗ = inf Jβ (f ). So there exists a matrix sequence {f k ∈ RN ×N : k = 1, 2, · · · }
f
such that lim Jβ (f k ) = J ∗ , which means α1 f k 2F ≤ Jβ (f k ) ≤ C0 for k =
k→∞
1, 2, · · · , i.e., f k 2F ≤ C0 /α1 . Therefore there exists a subsequence matrix of
{f k : k = 1, 2, · · · }, still denoted by {f k : k = 1, 2, · · · }, such that lim f k = f ∗ .
k→∞
Notice that |f |T V,β is also continuous with respect to f by (2.6), the conti-
nuity of Jβ (f ) with respect to f yields Jβ (f ∗ ) = lim Jβ (f k ) = J ∗ = inf Jβ (f ),
k→∞ f
i.e., f ∗ is the minimizer of Jβ (f ). The proof is complete.
Theorem 2. Denote by f † ∈ RN ×N the exact image. Then the minimizer f ∗ =

fα∗1 ,α2 ,β,δ satisfies the following estimates
√
P F T fα∗1 ,α2 ,β,δ F − P ĝ δ 2F ≤ δ 2 + α1 f † 2F + 2α2 N 2 β + 2α2 |f † |T V , (2.8)
δ2
√
fα∗1 ,α2 ,β,δ 2F ≤ 2α 1
+α 2 †
α1 |f |T V + α1 N
α2 2
β + f † 2F , (2.9)
∗ 2
† 2
√ †
|fα1 ,α2 ,β,δ |T V,β ≤ 2α2 + α2 f F + N β + |f |T V .
δ α1 2
(2.10)
Proof. Since fα∗1 ,α2 ,β,δ is the minimizer, we have
1
P F T fα∗1 ,α2 ,β,δ F − P ĝ δ 2F + α1 fα∗1 ,α2 ,β,δ 2F + α2 |fα∗1 ,α2 ,β,δ |T V,β
2
1
≤ P F T f † F − P ĝ δ 2F + α1 f † 2F + α2 |f † |T V,β
2
1 2
≤ δ + α1 f † 2F + α2 (|f † |T V,β − |f † |T V ) + α2 |f † |T V
2
1 N
β
= δ 2 + α1 f † 2F + α2 + α2 |f † |T V
2 † †
m,n=1 |fm,n |2 + β + |fm,n |2
1 2
≤ δ + α1 f † 2F + α2 N 2 β + α2 |f † |T V . (2.11)
2
Since |fα∗1 ,α2 ,β,δ |T V ≤ |fα∗1 ,α2 ,β,δ |T V,β , the proof is complete by the triangle
inequality.
The above decompositions are important for seeking the minimizer of our
cost functional, which is taken as our reconstruction of image. This result gen-
erates the resolution analysis for our reconstruction scheme in terms of the
data-matching and regularity-matching for the image, i.e., the quantitative error
descriptions on these two terms are given. We can adjust the parameters α1 , α2
analytically such that our reconstruction fits our concerns for either image details
(data-matching) or image sparsity (TV difference).
3 The Iteration Algorithm to Find the Minimizer

2
Take the image vector f = (f1 , f2 , · · · , fN 2 )T ∈ RN ×1 as the equivalent variables,
and each components fi has one-to-one correspondence relationship with fm,n ,
i.e. fm,n = f(n−1)N +m . For the optimization problem
1
min Jβ (f ) = min P F T f F − P ĝ δ 2F + α1 f 2F + α2 |f |T V,β (3.1)
f f 2
finding the minimizer f ∗ approximately, the Bregman iterative algorithm is given
in [13], which is established in terms of Bregman distance [14]. In order to solve
the optimization problem (3.1) iteratively, f (k+1) is yielded by solving its Euler-
Lagrange equation [15]. Due to the penalty terms f F and |f |T V,β , the corre-
sponding Euler-Lagrange equation for the minimizer is nonlinear. So we propose
8 X. Liu and J. Liu
to find the minimizer by the lagged diffusivity fixed point method [16]. Consid-
ering the optimization problem with respect to the image vector f as
1
min Jβ (f ) := min P(Ff ) − Pĝδ 22 + α1 f 22 + α2 |f |T V,β . (3.2)
f f 2
In order to solve the Euler-Lagrange equation of (3.1), we need the derivatives
of data matching term and the penalty terms in (3.2). By straightforward com-
putations, these derivatives have the following expressions:
⎧
⎨ ∇f 12 P F T f F − P ĝ δ 2F = F∗ P∗ PFf − F∗ P∗ Pĝδ ,
∇f f 2F = 2I ⊗ If , (3.3)
⎩
∇f |f |T V,β = L[f ]f ,
where P = I ⊗ P and the N 2 × N 2 matrix
L[f ] := (I ⊗ D)T Λ[f ](I ⊗ D) + (D ⊗ I)T Λ[f ](D ⊗ I) (3.4)
with
⎧
⎨ Λ[f ] := diag
d1 [f ] , · · · , dN 2 [f ] ,
1 1

⎩ d [f ] := (N 2 (I ⊗ D) f )2 + (N 2 (D ⊗ I) f )2 + β,
i l =1 i,l l l =1 i,l l
where i = i(m, n) = (n−1)N +m, l = l(m , n ) = (n −1)N +m for m, n, m , n =

1, · · · , N , and the N × N circulant matrix D := circulant(−1, 0, · · · , 0, 1),
Based on (3.3), we can find the approximate minimizer by the following
Bregman iterative algorithm.
Algorithm 1. Bregman iterative algorithm for minimizing Jβ (f )

δ N ×N
Input: frequency input {ĝm ,n : m , n = 1, · · · , N }, sampling matrix P ∈ R ,
and parameters α1 , α2 , β.
Do iteration from l = 0 with g (0) = Θ, f (0) = Θ.
While l < L0
{ Compute:
ĝ (l+1) = ĝ δ + ĝ (l) − F T f (l) F ,
f (l+1) = arg min α1 f 2F + α2 |f |T V,β + 12 P F T f F − P ĝ (l+1) 2F ,
f ∈RN ×N
l ⇐ l + 1. }
End do
f ∗ := f L0
End
According to (3.3), the stacking vector f (l+1) of the minimizer f (l+1) of Jβ (f )

at the l−th step satisfies the following nonlinear equation:

N 2 (I ⊗ P )f + 2α1 (I ⊗ I)f + α2 L[f ]f = F∗ I ⊗ P ĝδ − Ff (l) , (3.5)
with sampling data P ĝ δ and the spatial approximation f (l) at the (l − 1)th step,
which constitutes the standard Bregman iterative algorithm. Now we propose a
new algorithm based on the Bregman iteration by introducing an inner recursion.
Notice, the real symmetric matrix I ⊗ P may not be invertible due to our
finite sampling matrix P . Therefore an efficient algorithm should be developed
2
for solving the nonlinear system (3.5) with unknown f ∈ RN ×1 . We apply the
lagged diffusivity fixed point method [16].
1
Define Λn [f ] := diag( d(n−1)N +1 [f ]
, · · · , d(n−1)N
1
+N [f ]
), then Λ[f ] = diag(Λ1 [f ],
· · · , ΛN [f ]). Since
L1 [f ] := 2α1 (I ⊗ I) + α2 (I ⊗ D)T diag(Λ1 [f ], · · · , ΛN [f ])(I ⊗ D)
is a real positive block diagonal matrix and
L2 [f ] := α2 (D ⊗ I)T Λ[f ](D ⊗ I)
⎛ 1 ⎞
1
d1 [f ] + dN 2 [f ] − d11[f ] ··· 0 −d 1
N 2 [f ]
⎜ − d1 [f ] ⎟
d1 [f ] + d2 [f ] · · ·
1 1 1
⎜ 0 0 ⎟
⎜ . .. .. .. ⎟
= α2 ⎜⎜
.. .
..
. . . ⎟
⎟
⎜ ··· 1 1
−d 1 ⎟
⎝ 0 0 dN 2 −2 [f ] + dN 2 −1 [f ] N 2 −1 [f ]
⎠
− d 12 [f ] 0 ··· −d 1 1
+ 1
N N 2 −1 [f ] dN 2 −1 [f ] dN 2 [f ]
is a symmetric block matrix, we construct the inner iteration scheme from l-step
for the nonlinear system (3.5) as

L1 [f (l) ]f (l+1) = − N 2 (I ⊗ P ) + L2 [f (l) ] f (l)

+ I ⊗ P F∗ ĝδ − F∗ Ff (l) (3.6)
for l = 0, 1, · · · . Since L1 [f (l) ] is a known block diagonal matrix being symmetric

positive, the computational costs for solving f (l+1) are affordable by solving
each column vector of f (l+1) separately, which meets an N −dimensional linear
equations with symmetric positive coefficient matrix.
In the numerical experiments, we choose regularization of adjoint conjugate
gradient method (ACGM) as the inner iteration scheme to solve (3.6). Let b(l)
be the right term in (3.6) which is the known part from the l−step in exterior
recursion, and μ is the prior regularizing parameter in ACGM. So we have

(l+1) (l) (l) (l) (l)
fk+1 = fk − κk μfk + L1 [f (l) ]T (L1 [f (l) ]fk − b(l) ) , (3.7)
(l)
where κk is the step-size at k-step in inner recursion defining as
(l) (l)
(l) −rk , (μI + L1 [f (l) ]T L1 [f (l) ])(−rk )T
κk := (l)
,
(μI + L1 [f (l) ]T L1 [f (l) ])(−rk )22
from the classic successive over relaxation method (SOR), , is the L2 inner
(l) (l) (l)
product, and rk := μfk + L1 [f (l) ]T (L1 [f (l) ]fk − b(l) ).
10 X. Liu and J. Liu
(l)
Notice, the initial value f0 in the inner recursion can be chosen as 0 or f (l) ,
and the stopping criterion may be the maximum iteration number K0 or some
others related to the small values of the cost function or the small difference of
the iterative sequences. Here we stop the iteration process when the difference
between f (l+1) and f (l) is smaller than 10−3 . Finally we have the scheme to
find the approximate minimizer by the following iterative algorithm with inner
recursion.
Algorithm 2. Bregman iterative algorithm with inner recursion

δ N ×N
Input: frequency input {ĝm ,n : m , n = 1, · · · , N }, sampling matrix P ∈ R ,
and parameters α1 , α2 , β, μ, L0 .
Do exterior recursion from l = 0 with g (0) = Θ, f (0) = Θ
While l < L0
{ Compute: ĝ(l+1) = ĝδ + ĝ(l) − PFf (l) ,
(l)
Do inner recursion from k = 0 with f0 = f (l) .
(l+1) (l)
While fk+1 − fk 22 > 10−3
(l) (l)
{ Compute: b(l) , rk , κk .
(l+1) (l) (l) (l)
Compute: fk+1 = fk − κk rk with (3.7)
k ⇐ k + 1. }
End do
l ⇐ l + 1. }
End do
f ∗ := f L0
End
4 Numerical Experiments
All the numerical tests are performed in MATLAB 7.10 on a laptop with an
Intel Core i5 CPU M460 processor and 2 GB memory.
We consider a model problem with Ω = [0, 1]2 and N = 128. Define

2 2
1 1 1
D1 := x = (x1 , x2 ) : x1 − + x2 − ≤ ,
4 2 64

3 1 1 1
D2 := x = (x1 , x2 ) : x1 − ≤ , x2 − ≤ , (4.1)
4 8 2 4
and
⎧
⎪
⎨1, x ∈ D1 ,
f (x) := 2, x ∈ D2 , (4.2)
⎪
⎩
0, x ∈ Ω \ (D1 ∪ D2 ).
The functions f (x) together with its frequency function log(|fˆ(ω)|) in algorithm
domain (i.e., after shifting as Fig. 1) is shown in Fig. 2(a) and (b). Obviously, the
frequency data in the center (or in the four corners before shifting) consist of
the main information about the image, so we should sample these data as much
as possible.
Fig. 2. (a) f (x); (b) Frequency function log(|fˆ(ω)|) after shifting; (c) and (d) Sampling
noisy frequency data P60 log(|ĝ δ (ω)|) and P90 log(|ĝ δ (ω)|) after shifting.
δ
Firstly we yield the full noisy data gm,n from the exact image fm,n by
δ
gm,n = fm,n + δ × rand(m, n),
where m, n = 1, · · · , N and rand(m, n) are the random numbers in [−1, 1]. The
mesh image of initial image and the noisy image are shown in Fig. 3. Then the
full noisy frequency data are simulated by

,n = F[gm,n ], m , n = 1, · · · , N.
δ δ
ĝm (4.3)
So with the random band row sampling method using sampling matrix P ,
δ
P ĝm ,n is the input incomplete noisy data by row sampling process.
Fig. 3. The mesh of initial image f (x) and noisy image g δ (x).
Take α1 = 1000, α2 = 0.001, β = μ = 0.0001, and the noise level δ = 0.1. To

the row sampling processes, we consider two schemes by taking M0 = 60, cenR =
0.3 and M0 = 90, cenR = 0.3, so the sampling ratios are M0 /M = 60/128 =
46.88% and 70.31%, respectively. To compare the restoration performances by
applying more sampling data, we require that the data for M0 = 60 be included
in the data set for M0 = 90. In order to ensure the validity of tests, the random
number rand(m, n) and sampling rows should be fixed in each parts. Then we
obtain the sampling matrix P60 , P90 with pii = 1 only at the following locations:
i ∈ {1 − 5, 16, 17, 18, 23, 34, 37, 39, 40, 43, 44, 45, 47 − 55,
58, 60, 61, 63, 64, 70, 76 − 79, 81, 82, 83, 86, 88 − 91,
94, 95, 97, 98, 100, 101, 103, 105, 107 − 112, 125 − 128} (4.4)
12 X. Liu and J. Liu
and
i ∈ {1 − 14, 16, 17, 18, 23, 27 − 32, 34, 37, 39, 40, 43, 44, 45, 47 − 55,
58, 60, 61, 63, 64, 70, 72, 73, 74, 76 − 79, 81, 82, 83, 86, 88, 89 − 91,
94, 95, 97, 98, 100, 101, 103, 105, 107 − 112, 114 − 128} (4.5)
respectively. Figure 2(c) and (d) show the two-dimensional image of full fre-
quency data, the incomplete noisy frequency data with P60 , P90 after shifting
respectively.
In our iteration process, the Bregman iterative number is L0 = 20, and
the initial value in ACGM inner recursion is f (l) . We compare Algorithm 2
with Algorithm 1, i.e., comparing the proposed scheme to the Bregman iterative
algorithm without inner recursion. Figure 4(a) and (b) give the reconstructed
image f ∗ with P60 , P90 by our proposed algorithm, while Fig. 4(c) shows the
reconstructed image f ∗ with P90 by standard Bregman iterative algorithm.
Fig. 4. (a), (b) The reconstruction of f ∗ with P60 , P90 by our Bregman iterative algo-
rithm with ACGM inner recursion; (c) The reconstruction of f ∗ with P90 by standard
Bregman iterative algorithm without inner recursion.
From our numerical implementations, the algorithm based on random band

sampling method can reconstruct the piecewise smooth image with good edge-
preservation. Considering we apply the noisy data with relative error 10% and
the unused sampling data (the lost data) are more than 50% and 30%, the image
restorations based on Bregman iterative algorithm with ACGM inner recursion
are satisfactory. However, the reconstruction could only restore the relative grey
level in the whole image, the exact value cannot be recovered efficiently. The
numerical evidences for this phenomena are that the reconstructed image f ∗
with sampling matrix P90 has interfaces clearly, but the interfaces of f ∗ with
P60 is worse.
5 Conclusion
An efficient algorithm to restore image based on L2 − T V regularization penalty
terms is established. The data matching term of the optimizing model is only
used limited data in frequency domain, which are obtained by random band
sampling process. The new idea is that the model is included with two iteration:
Bregman iteration and adjoint conjugate gradient method as inner recursion.
In order to solve the optimizing problem, the Bregman iteration with lagged
Another random document with
no related content on Scribd:
the world. He has carried on that controversy since his Balliol days.
The exigencies that oblige him to pretend, against his better
knowledge and common civility, that I am petty and provincial and
patriotic and wilfully ignorant and pitifully out-of-date, oblige him to
pretend as much about most of those who stand for modern science
and a modern interpretation of history. He would pretend as hard
about Sir Ray Lankester, for example, or about Professor Gilbert
Murray or Sir Harry Johnston or Professor Barker, as he does about
me. It is a general system of pretence. It is a necessary part of—I will
not say of the Catholic attitude, but of his Catholic attitude towards
modern knowledge.
The necessity for a pose involving this pretence is not very
difficult to understand. Long before Mr. Belloc embarked upon the
present dispute he had become the slave of a tactical fiction, which
reiteration had made a reality for him. He evoked the fiction as early,
I believe, as his Oxford days. It may have been very effective at
Oxford—among the undergraduates. Then perhaps it was
consciously a defensive bluff, but certainly it is no longer that. He
has come at last to believe absolutely in this creature of his
imagination. He has come to believe this: that there is a vast
“modern European” culture of which the English-speaking world
knows nothing, of which the non-Catholic world knows nothing, and
with which he is familiar. It is on his side. It is always on his side. It is
simply and purely Belloccian. He certainly believes it is there. It
sustains his faith. It assuages the gnawing attacks of self-criticism
that must come to him in the night. Throughout these papers he is
constantly referring to this imaginary stuff—without ever coming to
precisions. Again and again and again and again—and again and
again and again, he alludes to this marvellous “European” science
and literature, beyond our ken.
He does not quote it; it does not exist for him to quote; but he
believes that it exists. He waves his hand impressively in the
direction in which it is supposed to be. It is his stand-by, his refuge,
his abiding fortress. But, in order to believe in it, it is necessary for
him to believe that no other English-speaking men can even read
French, and that their scepticism about it is based on some
“provincial” prejudice or some hatred of Catholics, or southern
people, or “Dagoes,” or “foreigners,” or what you will. That is why
Nature wilfully ignores the wonderful science of this “Europe”; and
why our Royal Society has no correspondence with it. But he has to
imagine it is there and make his readers imagine it is there, and that
there is this conspiracy of prejudice to ignore it, before he can even
begin to put up any appearance of a case against such a résumé of
current knowledge as the Outline of History.
Graceful Concessions to Mr. Belloc

All this rough and apparently irrelevant stuff about his own great
breadth and learning and my profound ignorance and provincialism,
to which he has devoted his two introductory papers, is therefore the
necessary prelude to putting over this delusion. That stream of
depreciation is not the wanton personal onslaught one might
suppose it to be at the first blush. If he has appeared to glorify
himself and belittle me, it is for greater controversial ends than a
mere personal score. We are dealing with a controversialist here and
a great apologist, and for all I know these may be quite legitimate
methods in this, to me, unfamiliar field.
Few people will be found to deny Mr. Belloc a considerable
amplitude of mind in his undertaking, so soon as they get thus far in
understanding him. Before he could even set about syndicating this
Companion to the Outline of History he had to incite a partisan
receptivity in the Catholic readers to whom he appeals, by declaring
that a violent hatred for their Church is the guiding motive of my life.
He had to ignore a considerable array of facts to do that, and he has
ignored them with great courage and steadfastness. He had to
arouse an indifferent Catholic public to a sense of urgent danger by
imposing this figure of a base, inveterate, and yet finally
contemptible enemy upon it. His is a greater task than mere dragon-
slaying. He had to create the dragon before he could become the
champion. And then, with his syndication arrangements complete,
while abusing me industriously for ignorance, backwardness, and
general intellectual backwoodism, he had to write the whole of these
articles without once really opening that Humbert safe of knowledge
which is his sole capital in this controversy. Time after time he refers
to it. Never once does he quote it. At most he may give us illusive
peeps....
Now and then as we proceed I shall note these illusive peeps.
I can admire great effort even when it is ill-directed, and to show
how little I bear him a grudge for the unpleasant things he has
induced himself to write about me, and for the still more unpleasant
things he tempts me—though I resist with a success that gratifies me
—to write about him, I contemplate a graceful compliment to Mr.
Belloc. In spite of the incurable ignorance of French and that “dirty
Dago” attitude towards foreigners Mr. Belloc has so agreeably put
upon me, it is my habit to spend a large part of the winter in a house
I lease among the olive terraces of Provence. There is a placard in
one corner of my study which could be rather amusingly covered
with the backs of dummy books. I propose to devote that to a
collection of Mr. Belloc’s authorities. There shall be one whole row at
least of the Bulletins of the Madame Humbert Society, and all the
later researches of the Belloc Academy of Anonymous Europeans,
bound in bluff leather. There will be Finis Darwinis by Hilario
Belloccio, and Hist. Eccles. by Hilarius Belloccius. I may have
occasion to refer to other leading authorities in the course of this
controversy. I shall add to it as we proceed.
And so, having examined, explained, disposed of, and in part
apologised for, Mr. Belloc’s personalities and the pervading
inelegance of his manners, I shall turn with some relief from this
unavoidably personal retort to questions of a more general interest. I
propose as my first study of these modern Catholic apologetics, so
valiantly produced by Mr. Belloc and so magnificently published and
displayed by the Catholic press, to follow our hero’s courageous but
unsteady progress through the mysteries of Natural Selection. And
after that we shall come to Original Sin and Human Origins in the
light of Mr. Belloc’s science and the phantom science of those
phantom naturalists and anthropologists he calls to his assistance.
II
THE THEORY OF NATURAL SELECTION
STATED
MY first article upon Mr. Belloc’s Companion to the Outline of History

dealt, much against my inclination and as charitably and amiably as
possible, with the oddities of Mr. Belloc’s manner and method, and
those remarkable non-existent “European authorities” to whom he
appeals habitually in moments of argumentative stress. I do not
propose to go on thus girding at Mr. Belloc. He is a Catholic
apologist, endorsed by Catholic authorities, and there is matter of
very great importance for our consideration in what he has to say
about the history of life and mankind.
After his second paper is finished his abuse of me becomes
merely incidental or indirect. He goes on to a staggering rush at
Natural Selection. Let us see to where Catholic thought has got—if
Mr. Belloc is to be trusted—in relation to this very fundamental
matter.
It is Mr. Belloc’s brilliant careless way to begin most of his
arguments somewhere about the middle and put the end first. His
opening peroration, so to speak, is a proclamation that this “Natural
Selection”—whatever it is—is “an old and done-for theory of Darwin
and Wallace.” It is “a laughing-stock for half a generation among
competent men.” Mr. George Bernard Shaw does not believe in it!
G. B. S. among the Fathers! That wonderful non-existent “latest
European work” which plays so large a part in Mr. Belloc’s dialectic is
summoned briefly, its adverse testimony is noted, and it is dismissed
to the safe again. And then there is a brief statement of how these
two vile fellows, Darwin and Wallace, set out upon this reprehensible
theorising. What a ruthless exposé it is of the true motives of
scientific people!
“The process of thought was as follows:

“‘There is no Mind at work in the universe; therefore changes
of this sort must come from blind chance or at least
mechanically. At all costs we must get rid of the idea of design:
of a desired end conceived in a Creative Mind. Here is a theory
which will make the whole process entirely mechanical and
dead, and get rid of the necessity for a Creator.’”
And so having invented and then as it were visited and spat

upon the derided and neglected tomb of Natural Selection and
assured us that God, Mr. Shaw, “European opinion,” and all good
Catholics are upon his side, Mr. Belloc plucks up courage and really
begins to write about Natural Selection.
Natural Selection is Pure Common-Sense

What is this Natural Selection which has been dead for half a
century, but which Mr. Belloc still exerts himself industriously through
four long papers to kill all over again? It is the purest common-sense,
the most obvious deduction from obvious facts. I have set out the
idea as plainly as I could in the Outline of History Mr. Belloc is
attacking. It is put so plainly there that, before he can begin to argue
against it, he has to misstate it; he has to tell the story all over again
in his own words and get it suitably askew. It was quite open to him
to quote from my account, but he preferred to compile his own
misstatement. Indeed, in all this argument against Natural Selection
he never once quotes my actual words. He paraphrases throughout.
He has put some words between inverted commas in one place, so
as inadvertently to produce the impression that they are mine, but
they are not mine.
Now the facts upon which the idea of Natural Selection rests are
matters of universal knowledge. “Every species of living thing is
continually dying and being born again as a multitude of fresh
individuals”; that is the primary fact. No species seems to be
perfectly adapted to its conditions, and even the happiest species
tends to multiply until it is in a state of need and pressure. So far
surely we are dealing with things beyond dispute. And next comes
the fact of individuality. Every living unit is individual with a difference
of its own. Every individual has its own distinctive differences, and
each of these differences may or may not be an advantage or a
disadvantage. Individuals with advantageous differences will
generally get on better in life, prosper and so be able to breed more
freely, than those with disadvantageous differences. Offspring have a
tendency to repeat the distinctive differences of their parents.
Therefore, taking a species as a whole by the million or billion or
million billion—for few species of animals or plants are represented
by fewer individuals than a million—there will be in each successive
generation a greater number of individuals with the differences that
are advantageous relative to the number with disadvantages. In
other words, the average of the species will have moved more or
less in the direction of the advantageous differences, whatever they
may be, and however numerous they may be. If, for example, the
species is chased and has to climb or run for it, there will be rather
more good climbers and sprinters in the new generation. There may
be other dangers and other needs; they will not affect the premium
set on quickness and the fate of the slow. And if the circumstances
of the species continue to press in the same direction, the movement
of the average will be in the same direction in this respect for so long
as they continue to press. Over a few score or even a few hundred
generations, and under conditions not very strenuous, a species may
not change very much. It may seem to be fixed in its general
characteristics, just as the continents seem to be fixed in their
general outline. But, as the range of time extends and the pressure
of necessity continues, the change becomes more striking.
Natural Selection Has Nothing to Do with
the Origin of Variations
That is the process of Natural Selection, the “laughing-stock” of
Mr. Belloc’s mysterious conclave of “European” savants. Natural
Selection has nothing to do with the reason for the differences
between individuals. It has no more to do with those than gravitation
has to do with the differences in the heaviness of different
substances. But it is necessary to state as much here, because in
some queer muddled way Mr. Belloc seems to be persuaded that it
has. These differences may arise by pure chance; they may come
about through the operation of complex laws, they may come in
shoals and have their seasons. These things have nothing to do with
Natural Selection.
Now, Wallace and Darwin were two excellent Europeans who
happened to be interested in natural history. In spite of the sinister
motives invented for them by Mr. Belloc, I doubt if any Catholic
sufficiently educated to have read their lives will agree that they had
even a latent animus against Catholic truth or even a subconscious
desire to “get rid of a Creator” in their minds. They no more thought
of “getting rid of a Creator” when humbly and industriously they
gathered their facts and put fact to fact than an honest bricklayer
thinks of “getting rid of a Creator” when he lays his bricks with care
and builds a sound piece of wall. They went about the world studying
natural history. They considered life with a patience and
thoroughness and freedom from preconceptions beyond the
imagination of a man of Mr. Belloc’s habits. They found no such
“fixity of species” as he is inspired to proclaim. They found much
evidence of a progressive change in species, and they saw no
reason to explain it by a resort to miracles or magic. A Catholic priest
of the Anglican communion named Malthus had written a very
interesting and suggestive book upon over-population and the
consequent struggle for existence between individuals. It turned the
attention of both these diligent and gifted observers to just that
process of Natural Selection I have stated. Independently both of
them came to the conclusions that species changed age by age and
without any necessary limits, and mainly through the sieve of Natural
Selection, and that, given a sufficient separation to reduce or prevent
interbreeding and a sufficient difference in the selective conditions at
work, two parts of the same species might change in different
directions, so as at last to become distinct and separate species.
Darwin’s book upon the subject was called The Origin of
Species. It was a very modest and sufficient title. He did not even go
to the length of calling it the origin of genera or orders or classes. He
did not at first apply it to man.
This is the theory of the origin of species through Natural
Selection. It was not pretended by either of these pioneers that
Natural Selection was the sole way through which the differences of
species came about. For example, Darwin devoted a considerable
part of his working life to such collateral modes of differentiation as
the hypothesis that Sexual Selection also had its share. Criticism has
whittled down that share to practically negligible proportions, but I
note the hypothesis here because it absolutely disposes of the
assertion which Mr. Belloc hammers on the table, that the Theory of
Natural Selection excludes any other modes of specific
differentiation.
Testing the Theory

Very rapidly this conception of Natural Selection was extended
by naturalists until it came to be regarded as the general process of
life. They came to realise that all species, all genera, all classes of
life, whatever else may be happening to them, are and always have
been varying through the process of Natural Selection, some rapidly,
some slowly; some so slowly as hardly to change at all through vast
ages. I have stated the a priori case by which, given birth and death
and individuality and changing conditions and sufficient time, it
appears logically inevitable that the change and differentiation of
species must occur, and must be now going on. If we had no
material evidence at all it would still be possible to infer the evolution
of species.
That a priori case has never been answered, and it seems to me
unanswerable. But scientific men, with their obstinate preference for
observation and experiment over mere logical gymnastics, rarely rest
their convictions on a priori cases. A sustaining scepticism is a
matter of conscience with them. To them an a priori case is merely a
theory—that is to say, a generalisation under trial. For nearly three-
quarters of a century, therefore, biologists have been examining
whatever instances they could discover that seemed to contradict
this assumption that the process of specific change under Natural
Selection is the general condition of life. To this day this view is still
called the Theory of Natural Selection, though to a great number it
has come to have the substantial quality of an embracing fact.
It would have been amusing if Mr. Belloc had told us more of his
ideas of the scientific world. Apparently he knows scarcely anything
of museums or laboratories or the spirit and methods of research.
And manifestly he has not the faintest suspicion of the way in which
the whole world of vital phenomena has been ransacked and
scrutinised to test, correct, supplement, amplify, or alter this great
generalisation about life. He probably shares the delusion of most
other men in the street, that scientific theories are scientific finalities,
that they are supposed to be as ultimate as the dogmas of some
infallible religion. He imagines them put over chiefly by asseveration,
just as the assertions of a polemical journalist are put over. He has
still to learn that theories are trial material, testing targets, directives
for research. Shooting at established theories is the normal
occupation of the scientific investigator. Mr. Belloc’s figure of the
scientific investigator is probably a queer, frowsty, and often, alas!
atheistical individual, poking about almost aimlessly among facts in
the hope of hitting upon some “discovery” or “getting rid of a God.”
He does not understand the tense relevance of the vast amount of
work in progress. But for three-quarters of a century the thought and
work of myriads of people round and about the world have borne
directly or almost directly upon the probing, sounding, testing, of the
theory of Natural Selection. It stands clarified and, it would seem,
impregnable to-day.
Some Irrelevant Questions

Among questions bearing upon it but not directly attacking it has
been the discussion of the individual difference. For example, are
differences due to individual experiences ever inherited? Or are only
inherent differences transmissible? What rôle is played by what one
might call “normal,” relatively slight differences, and what by the
“sports” and abnormal births in specific change? Do species under
stress, and feeding on strange food or living in unaccustomed
climates, betray any exceptional tendency to produce abnormality?
Have there been, so to speak, storms and riots of variation in some
cases? Can differences establish themselves while outer necessity
remains neutral? Can variations amounting to specific differences in
colour and form arise as a sort of play of the germ plasm and be
tolerated rather than selected by nature? In what manner do normal
differences arise? What happens to differences in cases of
hybridisation? Here are sample questions that have been the seeds
of splendid work and great arguments. Some of them were already
under discussion in Darwin’s time; he was a pioneer in such
explorations; many ideas of his have stood the test of time, and
many suggestions he threw out have been disproved. When some
casual “may be” of Darwin’s is examined and set aside, it is the
custom of polemical journalists to rush about and proclaim to all who
may be sufficiently ill-informed to listen that Darwin is “exploded.”
Such explosions of Darwin are constantly recurring like gun-fire near
a garrison town, and still he remains. None of these subsidiary
questions affect the stability of this main generalisation of biology,
the Theory of Natural Selection.
The actual attack and testing of the Theory of Natural Selection
have yielded negative results. The statement of the theory may have
been made finer and exacter, that is all. And yet the conditions of its
survival have been very exacting. If the theory is to stand, the whole
of plant and animal life in time and space must be arranged in a
certain order. It must be possible to replace classification by a
genealogical tree. Every form must fall without difficulty into its
proper place in that tree. If it is true that birds are descended from
reptiles or men from apes, then there must be no birds before the
reptiles appear, and no men before apes. The geological record is
manifestly a mere fragmentary history, still for the most part unread,
but, however fragmentary it is, it must be consistent. One human
skull in the coal measures blows the whole theory to atoms. The
passage from form to form must be explicable by intermediate types
capable of maintaining themselves; there may be gaps in the record,
but there must be no miraculous leaps in the story. If an animal living
in the air is to be considered as a lineal descendant of some animal
living in the water, then the structure of the former bit by bit and step
by step must be shown to be adapted, modified, changed about from
that of the latter; it must have ears for water-hearing modified for air-
hearing, and its heart and breathing arrangements must be shown to
be similarly changed over, and so on for all its structure. All these
requirements will follow naturally from the necessities of a process of
Natural Selection. They follow logically upon no other hypothesis.
They are not demanded, for example, by the idea of a Creator
continually interfering with and rectifying some stately,
unaccountable process of “Evolution,” which seems to be Mr.
Belloc’s idea—so far as he ventures to display any idea of his own—
in the matter. Such things as vestigial structures and a number of
odd clumsinesses in living things—many still very imperfect
adaptations to an erect position, for example—become grotesque in
relation to such a view. A Creator who put needless or inconvenient
fish structures into the anatomy of a land animal and made the whole
fauna and flora of the land a patch-up of aquatic forms of life must be
not so much a Divinity as a Pedant. But it is the burthen of the whole
beautiful science of comparative anatomy that the structure of
animals and plants, and their succession in time, fall exactly into the
conditions defined by the Theory of Natural Selection. In the most
lovely and intricate detail, in a vast multitude of examples, in plants
and in animals alike, this theme of the adaptation of pre-existing
structure is worked out.
We should in accordance with the Theory of Natural Selection
expect to find traces of the ancestral form, not only in the lay-out of
the adult animal, but in every phase of its life history, and that, in
fact, is just what we do find. There is no more fascinating branch of
comparative anatomy than embryology. Each life cycle we discuss
tends to repeat the ancestral story, and only under the stress of
necessity does it undergo modification at any point. There is little
toleration in the life process for unnecessary divergencies.
Economies are effected by short cuts and reductions, and special
fœtal structures are granted reluctantly. So that even in man we find
peeping through the adaptations imposed upon the human type by
its viviparous necessities, and in spite of the advantage of every
economy of force, memories, for example, of the gill slits, of the fish
heart and kidney, of the reptilian skull, of the mammalian tail. I
mention this fact in the Outline, and upon it Mr. Belloc comments in a
manner that leaves one’s doubts poised between his honesty and
his intelligence. He declares, which is totally untrue, that I “repeat the
old Victorian tag”—I doubt if there ever was such a tag—that the
embryo “climbs up the family tree.” He puts these words in inverted
commas as though I have really adopted and used them, and for the
life of me it is only by straining my charity to the utmost that I can
accept that this was an accident. Of course every text-book of
embryology for the last forty years has made it perfectly plain, as I
have stated here, that the life cycle can be and is modified at any
point, and that an embryo has much more serious work in hand than
reciting its family history. It betrays its ancestral origins to analysis;
but that is an altogether different matter. Mr. Belloc, however, is so
densely ignorant himself upon these questions that he can imagine,
or think it worth while to pretend to imagine and attempt to persuade
his readers by the expedient of these inverted commas, that I
entertain such a view. And then follow this, which I quote that the
reader may the better understand a certain occasional acerbity in my
allusions to Mr. Belloc:—
“He doesn’t know that Vailleton of Montpellier has knocked
the last nail into the coffin of that facile and superficial Victorian
shortcut (and blind alley). He has probably never heard of
Vailleton, and when he does he will suspect him for a foreigner.
That is what I mean by being provincial and not abreast of one’s
time.”
It is perfectly true that I have never heard of any Vailleton in

biological science. Nor has anyone else. There is “no sich person.”
Perhaps Mr. Belloc has not been able to read the manuscript of
some adviser, or his memory may have played a trick upon him.
Possibly he has in mind that eminent Victorian embryologist,
Vialleton, who, so far from being the very newest thing in “European”
biology, must now be getting on for seventy. He is half-way back to
Haeckel, the originator of the family-tree idea, a German
embryologist and not, as a matter of fact, the Victorian English
Protestant Mr. Belloc supposes him to be. Possibly years and years
ago some French student may have run away with the idea that
embryos conscientiously repeat their phylogeny, and Professor
Vialleton may have thought it well to discuss this idea in one of his
books. It is not an idea I have ever entertained, much less stated,
and its only interest here is that it gives Mr. Belloc a chance of
showing how rudely he can set out his inaccuracies and his
misconceptions.
But this is an incidental comment. I will reserve for my next
section a consideration of the remarkable arguments—“crushing
arguments” the enthusiastic cross-heads of his editor declare them
to be—that Mr. Belloc produces against this view of life as being in a
state of change under the action of Natural Selection, that I have put
here before the reader.
III
MR. BELLOC AS A SPECIMEN CRITIC OF
NATURAL SELECTION
THE chief arguments against the Theory of Natural Selection with

which Mr. Belloc has favoured us are neatly set out by him in two
triads. His passion for orderly arrangement is greater than his logic,
and we shall find that the second and third arguments of his second
triad are substantially the same. He is rather exceptionally ignorant
of modern scientific literature, and his arguments do not cover all the
countervailing considerations upon which systematic observation
and research work have been based—the speculations of Dr.
Fairfield Osborn would have been a godsend for him—but the things
he has to say are conveniently simple; they embody some prevalent
misconceptions, and they will be useful in accentuating the more
salient points in my account of the theory given in my second paper.
He produces first certain remarkable a priori arguments—his
“three a prior arguments.” The first is beautifully absurd. It is difficult
to believe it is advanced in anything but a spirit of burlesque. He
says that an advantage is not an advantage. He says that an
advantage does not give an advantage unless it is combined with
other advantages. You will think I am misrepresenting him. Then
please read this:—
“(1) The advantageous differences making for survival are

not of one kind in any particular case, but of an indefinitely large
number (e.g., climate getting colder needs not only warmer coat,
but power to digest new food, protective colouring so as not to
show dark against snow, etc., an indefinitely large number of
qualities). Now the chance of all being combined (and co-
ordinated) in a single individual, without design, accidentally, let
alone of their thus appearing in many individuals accidentally
and without design, approximates to zero.”
This is, so to speak, the short uncompleted form of the first

argument. It is expanded later to a copiousness too great to admit of
quotation. This expansion carries the statement right to its
conclusion, that only an individual possessing all the possible
differences that are advantageous at any particular time can survive.
Otherwise its differences have no “survival value.” They may be
advantages, but not sufficient advantages to score an advantage. I
know this sounds tipsy, but there it is in black and white in Mr.
Belloc’s wonderful Article V for any one to consult. It follows plainly
that, except for a miracle, every species must be exterminated in
every generation. I can see no other way out of it. No individual, he
declares, can survive without the full set of advantageous
differences, and the chance of any individual having the full set of
advantageous differences, he declares after some abstruse verbal
gestures, is zero. There is Mr. Belloc with his unfailing logic, his clear
mathematical demonstration, and all the rest of it. There is the lucid
Latin mind shining above my Nordic fog! Yet the previous generation
got along without any of the set! And species do survive.
Did Mr. Belloc imagine he was saying something else? It is not
for me to speculate. Helping out an antagonist in a controversy is apt
to be resented. He has, I think, simply got into a muddle here, and
he is not sufficiently self-critical to get out of it again. So he tries to
muddle through. It is quite reasonable to say that when a species is
under stress of changing conditions it is usual for the need for
adaptation to be felt upon a number of points and not simply upon
one, and that, since every advantage counts, the individuals with the
greatest combination of advantageous differences have the best
chances. But that does not alter the fact that even a single
advantage is an advantage. What happens in nature is not an
extermination of all who are not completely in the fashion of the new
differences. That seems to be Mr. Belloc’s idea, but it is a wrong
idea. What does happen is a diminution in each generation of the
number of the disadvantaged in relation to the number of the
advantaged. That is quite another affair. Mr. Belloc has not grasped
this. His third a priori argument shows as much even more plainly
than his first, and to that I shall presently come.
Mr. Belloc’s Mental Indigestion

I fancy this stuff he has written here is an outcome of an
indigestion of Samuel Butler by Mr. Belloc. I should not have thought
Mr. Belloc had read Samuel Butler, and I doubt if he has read him
much. But there is a decided echo of Luck or Cunning in the one
indistinct paragraph in which, without committing himself too deeply,
Mr. Belloc seems to convey his own attitude towards the procedure
of Evolution. “Design,” whatever that is, is at work, and Natural
Selection is not. “There is an innate power possessed by the living
thing to attempt its own adaptation.” It is quite a delusion apparently
that rabbits that cannot run or sparrows that are not quick on the
wing are killed off more frequently than the smarter fellows. That
never happens, though to the atheistically minded it may seem to
happen. If it happens, it would “get rid of a God.” But there are
rabbits which, unlike Mrs. Micawber, do make an effort. You must
understand that all creation, inspired by design, is striving. The good
fungus says to itself, “Redder and more spots will benefit me
greatly,” and tries and tries, and presently there are redder hues and
more spots. Or a happily inspired fish says: “There is a lot of food on
land and the life is more genteel there, so let me get lungs.” And
presently it gets lungs. Some day Mr. Belloc must take a holiday in
Sussex and flap about a bit and get himself some wings and
demonstrate all this. But perhaps this is caricature, and Mr. Belloc
when he talks about that “innate disposition” just means nothing very
much—just an attempt or something. I will not pretend to understand
Mr. Belloc fully upon this point.
Mr. Belloc’s Bird-Lizard
I will return to the essential misconception of the Theory of
Natural Selection betrayed in this first a priori when I consider Mr.
Belloc’s third feat of logic. But first let me glance at his second. In
this he says, very correctly, that every stage in the evolution of a
living creature must be a type capable of maintaining itself and every
change must be an advantageous change. I have noted this very
obvious point already in my second paper. But then Mr. Belloc
instructs us that the chances of its being so are, for no earthly
reason, zero—that fatal zero again!—and goes on to a passage so
supremely characteristic that it must be read to be believed:—
“A bird has wings with which it can escape its enemies. If it

began as a reptile without wings—when, presumably, it had
armour or some other aid to survival—what of the interval?
Natural Selection sets out to change a reptile’s leg into a bird’s
wing and the scales of its armour into feathers. It does so by
making the leg less and less of a leg for countless ages, and by
infinite minute gradations, gradually turning the scales into
feathers.
“By the very nature of the theory each stage in all these
millions is an advantage over the last towards survival! The thing
has only to be stated for its absurdity to appear. Compare the
‘get away’ chances of a lizard at one end of the process or a
sparrow at the other with some poor beast that had to try and
scurry off on half-wings! or to fly with half-legs!
“Postulate a design, say, ‘Here was something in the
making,’ and the process is explicable, especially if fairly rapid
so as to bridge over the dangerously weak stage of imperfection.
Postulate Natural Selection and it is manifestly impossible.”
Let us note a few things of which Mr. Belloc shows himself to be

unaware in this amusing display of perplexity. In the first place he
does not know that the Mesozoic reptiles most closely resembling
birds were creatures walking on their hind-legs, with a bony structure
of the loins and a backbone already suggestive of the avian
anatomy. Nor is he aware that in the lowliest of living birds the fore-
limbs are mere flappers, that the feathers are simpler in structure
than any other bird’s feathers, and that the general development of a
bird’s feather points plainly to the elongation of a scale. He has
never learnt that feathers came before wings, and that at first they
had to do, not with flying, but with protection against cold. Yet all this
was under his nose in the Outline of History in text and picture. The
transition from a quilled to a feathered dinosaur presents indeed no
imaginative difficulties, and the earliest birds ran and did not fly. One
of the earliest known extinct birds is Hesperornis, a wingless diving
bird. It is figured on page 30, and there is another bird on page 34
that Mr. Belloc might ponder with advantage. A whole great section
of living birds, like the ostrich and the emu, have no trace in their
structure of any ancestral flying phase; their breast-bones are
incapable of carrying the necessary muscular attachments.
H. G. Wells.
Low
But after the feather was fully developed it opened up great

possibilities of a strong and light extension of the flapper, helpful in
running or useful in leaps from tree to tree. Archæopteryx, another
early bird, which is also figured in the Outline, has a sort of bat-wing
fore-limb with feathers instead of membrane. It was a woodland
creature, and flew as a flying fox or a flying lemur or even a bat flies.
All these facts are widely known, and all that trouble about the half-
leg, half-wing, dissolves before them. But consider what a hash they
make of Mr. Belloc’s argument, and how pitifully it scurries off before
them on its nondescript stumps of pretentious half-knowledge, half-
impudence! So much for zero the second.
Troubles of Mr. Belloc as a Matrimonial

Agent
The final of this wonderful trinity of a prioris is a repetition of an
argument advanced ages ago by Queen Victoria’s Lord Salisbury,
when he was President of the British Association. Even then it struck
people that he had been poorly coached for the occasion. Assuming
that one or two individuals have got all these “survival value”
differences in the correct proportions—against which the chances
are zero—how by any theory of Natural Selection are we to suppose
they will meet, breed, and perpetuate them? So this argument runs.
The chances are again declared to be zero, the third zero, and Mr.
Belloc, I gather, calls in Design again here and makes his Creative
Spirit, which has already urged these two individuals, lions, or liver
flukes or fleas or what not, to make an effort and adapt themselves,
lead them now to their romantic and beneficial nuptials, while the
Theory of Natural Selection grinds its teeth in the background and
mutters “Foiled again.”
But this third argument reinforces the first, in showing what is the
matter with Mr. Belloc’s ideas in this group of questions. He has got
the whole business upside down. I rather blame the early Darwinians
in this matter for using so inaccurate a phrase as the “Survival of the
Fittest.” It is to that phrase that most of Mr. Belloc’s blunderings are
due. Yet he ought not to have been misled. He had a summary of
modern views before him. He criticises my Outline of History, he
abuses it, and yet he has an extraordinary trick of getting out of its
way whenever it swings near his brain-case. I warn the readers of
that modest compendium expressly (and as early as page 16) that
the juster phrase to use is not the Survival of the Fittest, but the
Survival of the Fitter. I do what I can throughout to make them see

Image and Graphics 9th International Conference ICIG 2017 Shanghai China September 13 15 2017 Revised Selected Papers Part III 1st Edition Yao Zhao

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Image and Graphics 9th International Conference ICIG 2017 Shanghai China September 13 15 2017 Revised Selected Papers Part III 1st Edition Yao Zhao

Uploaded by

Copyright:

Available Formats

Image and Graphics 9th International

Conference ICIG 2017 Shanghai China

Image and Graphics 10th International Conference ICIG

Image and Graphics 8th International Conference ICIG

Advances in Image and Graphics Technologies 12th

Smart Card Research and Advanced Applications 16th

Computational Intelligence and Intelligent Systems 9th

Image and Graphics 8th International Conference ICIG

Image Analysis and Processing ICIAP 2017 19th

Image Analysis and Processing ICIAP 2017 19th

David Taubman (Eds.)

ISSN 0302-9743 ISSN 1611-3349 (electronic)

Library of Congress Control Number: 2017960877

© Springer International Publishing AG 2017

Printed on acid-free paper

This Springer imprint is published by Springer Nature

Conference Place Date Submitted Proceeding

September 2017 Yao Zhao

Organizing Committee Chairs

Technical Program Chairs

Special Session Chairs

Technical Program Committee

Yan-Pei Cao Tsinghua University, China

Qi Li Peking University, China

Bin Wang Tsinghua University, China

Qiang Zhang Dalian University, China

A Double Recursion Algorithm to Image Restoration

RGB-D Saliency Detection with Multi-feature-fused Optimization . . . . . . . . . 15

Research on Color Image Segmentation . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

Depth-Based Focus Stacking with Labeled-Laplacian Propagation . . . . . . . . . 36

A Novel Method for Measuring Shield Tunnel Cross Sections . . . . . . . . . . . 47

Image Noise Estimation Based on Principal Component

Efficient High Dynamic Range Video Using Multi-exposure CNN Flow . . . . 70

Depth Image Acquisition Method in Virtual Interaction

An Image Segmentation Method Based on Asynchronous

Lagrange Detector in Image Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . 102

New Tikhonov Regularization for Blind Image Restoration . . . . . . . . . . . . . 113

Real-Time Multi-camera Video Stitching

Image Quality Assessment of Enriched Tonal Levels Images . . . . . . . . . . . . 134

Computer Graphics and Visualization

A Variational Model to Extract Texture from Noisy Image

Joint Visualization of UKF Tractography Data . . . . . . . . . . . . . . . . . . . . . . 158

Semantic Segmentation Based Automatic Two-Tone Portrait Synthesis . . . . . 170

Parameters Sharing Multi-items Non-parametric

SRG and RMSE-Based Automated Segmentation for Volume Data . . . . . . . . 194

Shape Recovery of Endoscopic Videos by Shape from Shading

Lazy Recoloring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214

Similar Trademark Image Retrieval Integrating LBP

Adaptive Learning Compressive Tracking Based on Kalman Filter . . . . . . . . 243

Online High-Accurate Calibration of RGB+3D-LiDAR

Run-Based Connected Components Labeling Using Double-Row Scan . . . . . 264

A 3D Tube-Object Centerline Extraction Algorithm

Moving Objects Detection in Video Sequences Captured

Fast Grid-Based Fluid Dynamics Simulation

Adaptive Density Optimization of Lattice Structures Sustaining

Hyperspectral Image Processing

Hyperspectral Image Classification Based on Deep Forest

Hyperspectral Image Classification Using Multi Vote Strategy

Efficient Deep Belief Network Based Hyperspectral Image Classification . . . . 347

Classification of Hyperspectral Imagery Based on Dictionary Learning

Deep Residual Convolutional Neural Network for Hyperspectral

Multi-view and Stereoscopic Processing

Stereoscopic Digital Camouflage Pattern Generation Algorithm

where i = i(m, n) = (n−1)N +m, l = l(m , n ) = (n −1)N +m for m, n, m , n =