Advances in Multimedia Information

Processing – PCM 2017: 18th

Pacific-Rim Conference on Multimedia,
Harbin, China, September 28-29, 2017,
Revised Selected Papers, Part II Bing
Zeng
Visit to download the full and correct content document:
https://textbookfull.com/product/advances-in-multimedia-information-processing-pcm-
2017-18th-pacific-rim-conference-on-multimedia-harbin-china-september-28-29-2017-
revised-selected-papers-part-ii-bing-zeng/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...

Biota Grow 2C gather 2C cook Loucas

https://textbookfull.com/product/biota-grow-2c-gather-2c-cook-
loucas/

Advances in Multimedia Information Processing – PCM

2018: 19th Pacific-Rim Conference on Multimedia, Hefei,
China, September 21-22, 2018, Proceedings, Part II
Richang Hong
https://textbookfull.com/product/advances-in-multimedia-
information-processing-pcm-2018-19th-pacific-rim-conference-on-
multimedia-hefei-china-september-21-22-2018-proceedings-part-ii-
richang-hong/

Advances in Multimedia Information Processing – PCM

2018: 19th Pacific-Rim Conference on Multimedia, Hefei,
China, September 21-22, 2018, Proceedings, Part III
Richang Hong
https://textbookfull.com/product/advances-in-multimedia-
information-processing-pcm-2018-19th-pacific-rim-conference-on-
multimedia-hefei-china-september-21-22-2018-proceedings-part-iii-
richang-hong/

Advances in Multimedia Information Processing PCM 2016

17th Pacific Rim Conference on Multimedia Xi an China
September 15 16 2016 Proceedings Part II 1st Edition
Enqing Chen
https://textbookfull.com/product/advances-in-multimedia-
information-processing-pcm-2016-17th-pacific-rim-conference-on-
multimedia-xi-an-china-september-15-16-2016-proceedings-part-
Advances in Multimedia Information Processing PCM 2016
17th Pacific Rim Conference on Multimedia Xi an China
September 15 16 2016 Proceedings Part I 1st Edition
Enqing Chen
https://textbookfull.com/product/advances-in-multimedia-
information-processing-pcm-2016-17th-pacific-rim-conference-on-
multimedia-xi-an-china-september-15-16-2016-proceedings-
part-i-1st-edition-enqing-chen/

Computational Linguistics and Intelligent Text

Processing 18th International Conference CICLing 2017
Budapest Hungary April 17 23 2017 Revised Selected
Papers Part II Alexander Gelbukh
https://textbookfull.com/product/computational-linguistics-and-
intelligent-text-processing-18th-international-conference-
cicling-2017-budapest-hungary-april-17-23-2017-revised-selected-
papers-part-ii-alexander-gelbukh/

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/

Image and Video Technology 8th Pacific Rim Symposium

PSIVT 2017 Wuhan China November 20 24 2017 Revised
Selected Papers 1st Edition Manoranjan Paul

https://textbookfull.com/product/image-and-video-technology-8th-
pacific-rim-symposium-psivt-2017-wuhan-china-
november-20-24-2017-revised-selected-papers-1st-edition-
manoranjan-paul/

Computational Linguistics and Intelligent Text

Processing 18th International Conference CICLing 2017
Budapest Hungary April 17 23 2017 Revised Selected
Papers Part I Alexander Gelbukh
https://textbookfull.com/product/computational-linguistics-and-
intelligent-text-processing-18th-international-conference-
cicling-2017-budapest-hungary-april-17-23-2017-revised-selected-
 Bing Zeng · Qingming Huang
Abdulmotaleb El Saddik · Hongliang Li
Shuqiang Jiang · Xiaopeng Fan (Eds.)

Advances in Multimedia
LNCS 10736

Information Processing –
PCM 2017
18th Pacific-Rim Conference on Multimedia
Harbin, China, September 28–29, 2017
Revised Selected Papers, Part II

123
Lecture Notes in Computer Science 10736
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, Chennai, 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/7409
Bing Zeng Qingming Huang
•

Abdulmotaleb El Saddik Hongliang Li

•

Shuqiang Jiang Xiaopeng Fan (Eds.)

•

Advances in Multimedia
Information Processing –
PCM 2017
18th Pacific-Rim Conference on Multimedia
Harbin, China, September 28–29, 2017
Revised Selected Papers, Part II

123
Editors
Bing Zeng Hongliang Li
University of Electronic Science University of Electronic Science
and Technology of China and Technology of China
Chengdu Chengdu
China China
Qingming Huang Shuqiang Jiang
University of Chinese Academy of Sciences Chinese Academy of Sciences
Beijing Beijing
China China
Abdulmotaleb El Saddik Xiaopeng Fan
University of Ottawa Harbin Institute of Technology
Ottawa, ON Harbin
Canada China

ISSN 0302-9743 ISSN 1611-3349 (electronic)

Lecture Notes in Computer Science
ISBN 978-3-319-77382-7 ISBN 978-3-319-77383-4 (eBook)
https://doi.org/10.1007/978-3-319-77383-4

Library of Congress Control Number: 2018935899

LNCS Sublibrary: SL3 – Information Systems and Applications, incl. Internet/Web, and HCI

© Springer International Publishing AG, part of Springer Nature 2018

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 the registered company Springer International Publishing AG
part of Springer Nature
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
 Preface

On behalf of the Organizing Committee, it is our great pleasure to welcome you to the
proceedings of the 2017 Pacific-Rim Conference on Multimedia (PCM 2017). PCM
serves as an international forum to bring together researchers and practitioners from
academia and industry to discuss research on state-of-the-art Internet multimedia
processing, multimedia service, analysis, and applications. PCM 2017 was the 18th in
the series that has been held annually since 2000. In 2017, PCM was held in Harbin,
China.
Consistent with previous editions of PCM, we prepared a very attractive technical
program with two keynote talks, one best paper candidate session, nine oral presen-
tation sessions, two poster sessions, and six oral special sessions. Moreover, thanks to
the co-organization with IEEE CAS Beijing chapter, this year’s program featured a
panel session titled “Advanced Multimedia Technology.” Social and intellectual
interactions were enjoyed among students, young researchers, and leading scholars.
We received 264 submissions for regular papers this year. These submissions cover
the areas of multimedia content analysis, multimedia signal processing and systems,
multimedia applications and services, etc. We thank our 104 Technical Program
Committee members for their efforts in reviewing papers and providing valuable
feedback to the authors. From the total of 264 submissions and based on at least two
reviews per submission, the Program Chairs decided to accept 48 oral papers (18.2%)
and 96 poster papers, i.e, the overall acceptance ratio for regular paper is 54.9%.
Among the 48 oral papers, two papers received the Best Paper and the Best Student
Paper award. Moreover, we accepted six special sessions with 35 papers.
The technical program is an important aspect but only delivers its full impact if
surrounded by challenging keynotes. We are extremely pleased and grateful to have
two exceptional keynote speakers, Wenwu Zhu and Josep Lladós, accept our invitation
and present interesting ideas and insights at PCM 2017. We would also like to express
our sincere gratitude to all the other Organizing Committee members, the general
chairs, Bing Zeng, Qingming Huang, and Abdulmotaleb El Saddik, the program chair,
Hongliang Li, Shuqiang Jiang, and Xiaopeng Fan, the panel chairs, Zhu Li and Debin
Zhao, the organizing chairs, Shaohui Liu, Liang Li, and Yan Chen, the publication
chairs, Shuhui Wang and Wen-Huang Cheng, the sponsorship chairs, Wangmeng Zuo,
Luhong Liang, and Ke Lv, the registration and finance chairs, Guorong Li and Weiqing
Min, among others. Their outstanding effort contributed to this extremely rich and
complex main program that characterizes PCM 2017. Last but not the least, we thank
VI Preface

all the authors, session chairs, student volunteers, and supporters. Their contributions
are much appreciated.
We sincerely hope that you will enjoy reading the proceedings of PCM 2017.

September 2017 Bing Zeng

Qingming Huang
Abdulmotaleb El Saddik
Hongliang Li
Shuqiang Jiang
Xiaopeng Fan
 Organization

Organizing Committee
General Chairs
Bing Zeng University of Electronic Science and Technology
of China
Qingming Huang University of Chinese Academy of Sciences, China
Abdulmotaleb El Saddik University of Ottawa, Canada

Program Chairs
Hongliang Li University of Electronic Science and Technology
of China
Shuqiang Jiang ICT, Chinese Academy of Sciences, China
Xiaopeng Fan Harbin Institute of Technology, China

Organizing Chairs
Shaohui Liu Harbin Institute of Technology, China
Liang Li University of Chinese Academy Sciences, China
Yan Chen University of Electronic Science and Technology
of China

Panel Chairs
Zhu Li University of Missouri-Kansas City, USA
Debin Zhao Harbin Institute of Technology, China

Technical Committee
Publication Chairs
Shuhui Wang ICT, Chinese Academy of Sciences, China
Wen-Huang Cheng Taiwan Academia Sinica, Taiwan
Tongwei Ren Nanjing University, China
Lu Fang Hong Kong University of Science and Technology,
SAR China

Special Session Chairs

Yan Liu The Hong Kong Polytechnic University, SAR China
Yu-Gang Jiang Fudan University, China
Wen Ji ICT, Chinese Academy of Sciences, China
Jinqiao Shi Chinese Academy Sciences, China
Feng Jiang Harbin Institute of Technology, China
VIII Organization

Tutorial Chairs
Zheng-jun Zha Hefei Institute of Intelligent Machines,
Chinese Academy of Sciences, China
Siwei Ma Peking University, China
Chong-Wah Ngo City University of Hong Kong, SAR China
Ruiqin Xiong Peking University, China

Publicity Chairs
Liang Lin Sun Yat-sen University, China
Luis Herranz Computer Vision Center, Spain
Cees Snoek University of Amsterdam and Qualcomm Research,
The Netherlands
Shin’ichi Satoh National Institute of Informatics, Japan
Zi Huang The University of Queensland, Australia

Sponsorship Chairs
Wangmeng Zuo Harbin Institute of Technology, China
Luhong Liang ASTRI, Hong Kong, SAR China
Ke Lv University of Chinese Academy of Sciences, China

Registration Chairs
Guorong Li University of Chinese Academy of Sciences, China
Shuyuan Zhu University of Electronic Science and Technology
of China
Wenbin Yin Harbin Institute of Technology, China

Finance Chairs
Weiqing Min ICT, Chinese Academy of Sciences, China
Wenbin Che Harbin Institute of Technology, China
 Contents – Part II

Content Analysis

A Competitive Combat Strategy and Tactics in RTS Games AI

and StarCraft . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Adil Khan, Kai Yang, Yunsheng Fu, Fang Lou, Worku Jifara,
Feng Jiang, and Liu Shaohui

Indoor Scene Classification by Incorporating Predicted Depth Descriptor . . . . 13

Yingbin Zheng, Jian Pu, Hong Wang, and Hao Ye

Multiple Thermal Face Detection in Unconstrained Environments

Using Fully Convolutional Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
Yezhao Fan, Guangtao Zhai, Jia Wang, Menghan Hu, and Jing Liu

Object Proposal via Depth Connectivity Constrained Grouping . . . . . . . . . . . 34

Yuantian Wang, Lei Huang, Tongwei Ren, Sheng-Hua Zhong,
Yan Liu, and Gangshan Wu

Edge-Aware Saliency Detection via Novel Graph Model . . . . . . . . . . . . . . . 45

Hanpei Yang and Weihai Li

Multiple Kernel Learning Based on Weak Learner for Automatic

Image Annotation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
Hua Zhong, Xu Yuan, Zhikui Chen, Fangming Zhong, and Yonglin Leng

An Efficient Feature Selection for SAR Target Classification . . . . . . . . . . . . 68

Moussa Amrani, Kai Yang, Dongyang Zhao, Xiaopeng Fan,
and Feng Jiang

Fine-Art Painting Classification via Two-Channel Deep Residual Network . . . 79

Xingsheng Huang, Sheng-hua Zhong, and Zhijiao Xiao

Automatic Foreground Seeds Discovery for Robust Video

Saliency Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
Lin Zhang, Yao Lu, and Tianfei Zhou

Semantic R-CNN for Natural Language Object Detection. . . . . . . . . . . . . . . 98

Shuxiong Ye, Zheng Qin, Kaiping Xu, Kai Huang, and Guolong Wang

Spatio-Temporal Context Networks for Video Question Answering . . . . . . . . 108

Kun Gao and Yahong Han
X Contents – Part II

Object Discovery and Cosegmentation Based on Dense Correspondences. . . . 119

Yasi Wang, Hongxun Yao, Wei Yu, and Xiaoshuai Sun

Semantic Segmentation Using Fully Convolutional Networks and Random

Walk with Prediction Prior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
Xiaoyu Lei, Yao Lu, Tingxi Liu, and Xiaoxue Shi

Multi-modality Fusion Network for Action Recognition . . . . . . . . . . . . . . . . 139

Kai Huang, Zheng Qin, Kaiping Xu, Shuxiong Ye, and Guolong Wang

Fusing Appearance Features and Correlation Features for Face

Video Retrieval. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
Chenchen Jing, Zhen Dong, Mingtao Pei, and Yunde Jia

A Robust Image Reflection Separation Method Based on Sift-Edge Flow. . . . 161

Shaomin Du, Xiaohui Liang, and Xiaochuan Wang

A Fine-Grained Filtered Viewpoint Informed Keypoint Prediction

from 2D Images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
Qingnan Li, Ruimin Hu, Yixin Chen, Jingwen Yan, and Jing Xiao

More Efficient, Adaptive and Stable, A Virtual Fitting System

Using Kinect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
Chang-Tai Xiong, Shun-Lei Tang, and Ruo-Yu Yang

Exploiting Sub-region Deep Features for Specific Action Recognition

in Combat Sports Video. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
Yongqiang Kong, Zhaoqiang Wei, Zhengang Wei, Shengke Wang,
and Feng Gao

Face Anti-spoofing Based on Motion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202

Ran Wang, Jing Xiao, Ruimin Hu, and Xu Wang

A Novel Action Recognition Scheme Based on Spatial-Temporal

Pyramid Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
Hengying Zhao and Xinguang Xiang

Co-saliency Detection via Sparse Reconstruction and Co-salient

Object Discovery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
Bo Li, Zhengxing Sun, Jiagao Hu, and Junfeng Xu

Robust Local Effective Matching Model for Multi-target Tracking . . . . . . . . 233

Hao Sheng, Li Hao, Jiahui Chen, Yang Zhang, and Wei Ke

Group Burstiness Weighting for Image Retrieval. . . . . . . . . . . . . . . . . . . . . 244

Mao Wang, Qiang Liu, Yuewei Ming, and Jianping Yin
 Contents – Part II XI

Stereo Saliency Analysis Based on Disparity Influence

and Spatial Dissimilarity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254
Lijuan Duan, Fangfang Liang, Wei Ma, and Shuo Qiu

Object Classification of Remote Sensing Images Based

on Rotation-Invariant Discrete Hashing . . . . . . . . . . . . . . . . . . . . . . . . . . . 264
Hui Xu, Yazhou Liu, and Quansen Sun

Robust Principal Component Analysis via Symmetric Alternating

Direction for Moving Object Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . 275
Zhenzhou Shao, Gaoyu Wu, Ying Qu, Zhiping Shi, Yong Guan,
and Jindong Tan

Driver Head Analysis Based on Deeply Supervised Transfer Metric

Learning with Virtual Data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286
Keke Liu, Yazhou Liu, Quansen Sun, Sugiri Pranata, and Shengmei Shen

Joint Dictionary Learning via Split Bregman Iteration for Large-Scale

Image Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 296
Yanyun Qu, Hanqian Li, and Yan Zhang

Multi-operator Image Retargeting with Preserving Aspect Ratio

of Important Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 306
Qian Zhang, Zhenhua Tang, Hongbo Jiang, and Kan Chang

Human Action Recognition in Videos of Realistic Scenes Based

on Multi-scale CNN Feature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 316
Yongsheng Zhou, Nan Pu, Li Qian, Song Wu, and Guoqiang Xiao

Automatic Facial Complexion Classification Based on Mixture Model. . . . . . 327

Minjie Xu, Chunrong Guo, Yangyang Hu, Hong Lu, Xue Li, Fufeng Li,
and Wenqiang Zhang

Spectral Context Matching for Video Object Segmentation

Under Occlusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337
Xiaoxue Shi, Yao Lu, Tianfei Zhou, and Xiaoyu Lei

Hierarchical Tree Representation Based Face Clustering

for Video Retrieval . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 347
Pengyi Hao, Edwin Manhando, Cong Bai, and Yujiao Huang

Improved Key Poses Model for Skeleton-Based Action Recognition . . . . . . . 358

Xiaoqiang Li, Yi Zhang, and Junhui Zhang

Pic2Geom: A Fast Rendering Algorithm for Low-Poly Geometric Art . . . . . . 368

Ruisheng Ng, Lai-Kuan Wong, and John See
XII Contents – Part II

Attention Window Aware Encoder-Decoder Model for Spoken

Language Understanding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 378
Yiming Wang, Wenge Rong, Jingshuang Liu, Jingfei Han,
and Zhang Xiong

A New Fast Algorithm for Sample Adaptive Offset. . . . . . . . . . . . . . . . . . . 388

Chentian Sun, Yang Wang, Xiaopeng Fan, and Debin Zhao

Motion-Compensated Deinterlacing Based on Scene Change Detection . . . . . 397

Xiaotao Zhu, Qian Huang, Feng Ye, Fan Liu, Shufang Xu,
and Yanfang Wang

Center-Adaptive Weighted Binary K-means for Image Clustering . . . . . . . . . 407

Yinhe Lan, Zhenyu Weng, and Yuesheng Zhu

Aligned Local Descriptors and Hierarchical Global Features

for Person Re-Identification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 418
Yihao Zhang, Wenmin Wang, and Jinzhuo Wang

A Novel Background Subtraction Method Based on ViBe . . . . . . . . . . . . . . 428

Jian Liao, Hanzi Wang, Yan Yan, and Jin Zheng

Layout-Driven Top-Down Saliency Detection for Webpage . . . . . . . . . . . . . 438

Xixi Li, Di Liu, Kao Zhang, and Zhenzhong Chen

Saliency Detection by Superpixel-Based Sparse Representation . . . . . . . . . . . 447

Guangyao Chen and Zhenzhong Chen

Reading Two Digital Video Clocks for Broadcast Basketball Videos . . . . . . . 457
Xinguo Yu, Xiaopan Lyu, Lei Xiang, and Hon Wai Leong

Don’t Be Confused: Region Mapping Based Visual Place Recognition . . . . . 467

Dapeng Du, Na Liu, Xiangyang Xu, and Gangshan Wu

An Effective Head Detection Framework via Convolutional

Neural Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 477
Canmiao Fu, Yule Yuan, Qiang Zeng, Siying He, and Yong Zhao

Identifying Gambling and Porn Websites with Image Recognition. . . . . . . . . 488

Longxi Li, Gaopeng Gou, Gang Xiong, Zigang Cao, and Zhen Li

Image-Set Based Collaborative Representation for Face Recognition

in Videos . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 498
Gaopeng Gou, Junzheng Shi, Gang Xiong, Peipei Fu, Zhen Li,
and Zhenzhen Li

Vectorized Data Combination and Binary Search Oriented Reweight

for CPU-GPU Based Real-Time 3D Ball Tracking . . . . . . . . . . . . . . . . . . . 508
Ziwei Deng, Yilin Hou, Xina Cheng, and Takeshi Ikenaga
 Contents – Part II XIII

Hot Topic Trend Prediction of Topic Based on Markov Chain

and Dynamic Backtracking. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 517
Feng Xu, Jue Liu, Ying He, and Yating Hou

Fast Circular Object Localization and Pose Estimation for Robotic

Bin Picking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 529
Linyao Luo, Yanfei Luo, Hong Lu, Haowei Yuan, Xuehua Tang,
and Wenqiang Zhang

Local Temporal Coherence for Object-Aware Keypoint Selection

in Video Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 539
Songlin Du and Takeshi Ikenaga

A Combined Feature Approach for Speaker Segmentation

Using Convolution Neural Network. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 550
Jiang Zhong, Pan Zhang, and Xue Li

DDSH: Deep Distribution-Separating Hashing for Image Retrieval . . . . . . . . 560

Junjie Chen and Anran Wang

An Obstacle Detection Method Based on Binocular Stereovision . . . . . . . . . 571

Yihan Sun, Libo Zhang, Jiaxu Leng, Tiejian Luo, and Yanjun Wu

Coding, Compression, Transmission, and Processing

Target Depth Measurement for Machine Monocular Vision . . . . . . . . . . . . . 583

Jiafa Mao, Mingguo Zhang, Linan Zhu, Cong Bai, and Gang Xiao

Automatic Background Adjustment for Chinese Paintings

Using Pigment Lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 596
Jie Guo, Chunyou Li, and Jingui Pan

Content-Based Image Recovery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 606

Hong-Yu Zhou and Jianxin Wu

Integrating Visual Word Embeddings into Translation Language Model

for Keyword Spotting on Historical Mongolian Document Images. . . . . . . . . 616
Hongxi Wei, Hui Zhang, and Guanglai Gao

The Analysis for Binaural Signal’s Characteristics of a Real Source

and Corresponding Virtual Sound Image . . . . . . . . . . . . . . . . . . . . . . . . . . 626
Jinshan Wang, Xiaochen Wang, Weiping Tu, Jun Chen, Tingzhao Wu,
and Shanfa Ke

Primary-Ambient Extraction Based on Channel Pair for 5.1 Channel

Audio Using Least Square . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 634
Dingyan Song, Ge Gao, Yi Chen, and Xi Hu
XIV Contents – Part II

Multi-scale Similarity Enhanced Guided Normal Filtering . . . . . . . . . . . . . . 645

Wenbo Zhao, Xianming Liu, Shiqi Wang, and Debin Zhao

Deep Residual Convolution Neural Network for Single-Image

Robust Crowd Counting. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 654
Mingjie Lu and Bo Yan

An Efficient Method Using the Parameterized HRTFs for 3D Audio

Real-Time Rendering on Mobile Devices . . . . . . . . . . . . . . . . . . . . . . . . . . 663
Yucheng Song, Weiping Tu, Ruimin Hu, Xiaochen Wang, Wei Chen,
and Cheng Yang

Efficient Logo Insertion Method for High-Resolution H.265/HEVC

Compressed Video . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 674
Qi Jing, Peng Xu, Jun Sun, and Zongming Guo

Image Decomposition Based Nighttime Image Enhancement . . . . . . . . . . . . 683

Xuesong Jiang, Hongxun Yao, and Dilin Liu

PSNR Estimate for JPEG Compression . . . . . . . . . . . . . . . . . . . . . . . . . . . 693

Ci Wang, Ying Yang, and Jianhua Shen

Speech Intelligibility Enhancement in Strong Mechanical Noise Based

on Neural Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 702
Feng Cheng, Xiaochen Wang, Li Gang, Weiping Tu, and Jinshan Wang

Interactive Temporal Visualization of Collaboration Networks . . . . . . . . . . . 713

Ming Jing, Xueqing Li, and Yupeng Hu

On the Impact of Environmental Sound on Perceived Visual Quality. . . . . . . 723

Wenhan Zhu, Guangtao Zhai, Wei Sun, Yi Xu, Jing Liu, Yucheng Zhu,
and Xiaokang Yang

A Novel Texture Exemplars Extraction Approach Based on Patches

Homogeneity and Defect Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 735
Hui Lai, Lulu Yin, Huisi Wu, and Zhenkun Wen

Repetitiveness Metric of Exemplar for Texture Synthesis . . . . . . . . . . . . . . . 745

Lulu Yin, Hui Lai, Huisi Wu, and Zhenkun Wen

Unsupervised Cross-Modal Hashing with Soft Constraint . . . . . . . . . . . . . . . 756

Yuxuan Zhou, Yaoxian Li, Rui Liu, Lingyun Hao, and Yuanliang Sun

Scalable Video Coding Based on the User’s View for Real-Time Virtual
Reality Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 766
Hao Jiang, Gang He, Wenxin Yu, Zheng Wang, and Yunsong Li
 Contents – Part II XV

Towards Visual SLAM with Memory Management

for Large-Scale Environments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 776
Fu Li, Shaowu Yang, Xiaodong Yi, and Xuejun Yang

Entropy Based Sub-band Deletion for Multispectral Image Compression . . . . 787

Worku J. Sori, Zhao Dongyang, Lou Fang, Fu Yunsheng, Liu Shaohui,
Feng Jiang, and Khan Adil

Automatic Texture Exemplar Extraction Based on a Novel

Textureness Metric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 798
Huisi Wu, Junrong Jiang, Ping Li, and Zhenkun Wen

In Defense of Fully Connected Layers in Visual Representation Transfer . . . . 807

Chen-Lin Zhang, Jian-Hao Luo, Xiu-Shen Wei, and Jianxin Wu

Block Cluster Based Dictionary Learning for Image De-noising

and De-blurring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 818
JianWei Zheng, Ping Yang, Shanshan Fang, and Cong Bai

Content Adaptive Constraint Based Image Upsampling . . . . . . . . . . . . . . . . 827

Fan Yang, Huizhu Jia, Don Xie, Rui Chen, and Wen Gao

Image Quality Assessment for Video Surveillance System . . . . . . . . . . . . . . 838

Jianhua Shen, Hongyan Zhang, and Ci Wang

Style Transfer Based on Style Primitive Discovery . . . . . . . . . . . . . . . . . . . 847

Hao Wu, Zhengxing Sun, Shuang Wang, Weihang Yuan,
and Hui-Hsia Chen

Construction of Sampling Two-Channel Nonseparable Wavelet Filter Bank

and Its Fusion Application for Multispectral Image Pansharpening . . . . . . . . 859
Bin Liu, Weijie Liu, and Longxiang Xu

Data Reconstruction Based on Supervised Deep Auto-Encoder . . . . . . . . . . . 869

Ting Rui, Sai Zhang, Tongwei Ren, Jian Tang, and Junhua Zou

A Novel Fragile Watermarking Scheme for 2D Vector

Map Authentication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 880
Guoyin Zhang, Qingan Da, Liguo Zhang, Jianguo Sun, Qilong Han,
Liang Kou, and WenShan Wang

Hybrid Domain Encryption Method of Hyperspectral Remote

Sensing Image . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 890
Wenhao Geng, Jing Zhang, Lu Chen, Jiafeng Li, and Li Zhuo

Anomaly Detection with Passive Aggressive Online Gaussian

Model Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 900
Zheran Hong, Bin Liu, and Nenghai Yu
XVI Contents – Part II

Multi-scale Convolutional Neural Networks for Non-blind

Image Deconvolution. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 911
Xuehui Wang, Feng Dai, Jinli Suo, Yongdong Zhang, and Qionghai Dai

Feature-Preserving Mesh Denoising Based on Guided Normal Filtering . . . . . 920

Renjie Wang, Wenbo Zhao, Shaohui Liu, Debin Zhao, and Chun Liu

Visual-Inertial RGB-D SLAM for Mobile Augmented Reality . . . . . . . . . . . 928

Williem, Andre Ivan, Hochang Seok, Jongwoo Lim, Kuk-Jin Yoon,
Ikhwan Cho, and In Kyu Park

ODD: An Algorithm of Online Directional Dictionary Learning

for Sparse Representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 939
Dan Xu, Xinwei Gao, Xiaopeng Fan, Debin Zhao, and Wen Gao

A Low Energy Multi-hop Routing Protocol Based on Programming Tree

for Large-Scale WSN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 948
Feng Xu, Yating Hou, Guozhong Qian, and Yunyu Yao

Sparse Stochastic Online AUC Optimization for Imbalanced

Streaming Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 960
Min Yang, Xufen Cai, Ruimin Hu, Long Ye, and Rong Zhu

Traffic Congestion Level Prediction Based on Video

Processing Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 970
Wenyu Xu, Guogui Yang, Fu Li, and Yuanhang Yang

Coarse-to-Fine Multi-camera Network Topology Estimation . . . . . . . . . . . . . 981

Chang Xing, Sichen Bai, Yi Zhou, Zhong Zhou, and Wei Wu

An Adaptive Tuning Sparse Fast Fourier Transform . . . . . . . . . . . . . . . . . . 991

Sheng Shi, Runkai Yang, Xinfeng Zhang, Haihang You,
and Dongrui Fan

Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1001

 Contents – Part I

Best Paper Candidate

Deep Graph Laplacian Hashing for Image Retrieval . . . . . . . . . . . . . . . . . . 3

Jiancong Ge, Xueliang Liu, Richang Hong, Jie Shao,
and Meng Wang

Deep Video Dehazing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

Wenqi Ren and Xiaochun Cao

Image Tagging by Joint Deep Visual-Semantic Propagation . . . . . . . . . . . . . 25

Yuexin Ma, Xinge Zhu, Yujing Sun, and Bingzheng Yan

Exploiting Time and Frequency Diversities for High-Quality

Linear Video Transmission: A MCast Framework . . . . . . . . . . . . . . . . . . . . 36
Chaofan He, Huiying Wang, Yang Hu, Yan Chen,
and Houqiang Li

Light Field Image Compression with Sub-apertures Reordering

and Adaptive Reconstruction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
Chuanmin Jia, Yekang Yang, Xinfeng Zhang, Shiqi Wang,
Shanshe Wang, and Siwei Ma

Video Coding

Fast QTBT Partition Algorithm for JVET Intra Coding

Based on CNN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
Zhipeng Jin, Ping An, and Liquan Shen

A Novel Saliency Based Bit Allocation and RDO for HEVC . . . . . . . . . . . . 70

Jiajun Xu, Qiang Peng, Bing Wang, Changbin Li,
and Xiao Wu

Light Field Image Compression Scheme Based on MVD

Coding Standard . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
Xinpeng Huang, Ping An, Liquan Shen, and Kai Li

A Real-Time Multi-view AVS2 Decoder on Mobile Phone . . . . . . . . . . . . . 89

Yingfan Zhang, Zhenan Lin, Weilun Feng, Jun Sun,
and Zongming Guo
XVIII Contents – Part I

Compressive Sensing Depth Video Coding via Gaussian Mixture

Models and Object Edges. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
Kang Wang, Xuguang Lan, Xiangwei Li, Meng Yang,
and Nanning Zheng

Image Super-Resolution, Debluring, and Dehazing

AWCR: Adaptive and Weighted Collaborative Representations

for Face Super-Resolution with Context Residual-Learning. . . . . . . . . . . . . . 107
Tao Lu, Lanlan Pan, Jiaming Wang, Yanduo Zhang,
Zhongyuan Wang, and Zixiang Xiong

Single Image Haze Removal Based on Global-Local Optimization

for Depth Map . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
Hongda Zhang, Yuanyuan Gao, Hai-Miao Hu, Qiang Guo,
and Yukun Cui

Single Image Dehazing Using Deep Convolution Neural Networks . . . . . . . . 128

Shengdong Zhang, Fazhi He, and Jian Yao

SPOS: Deblur Image by Using Sparsity Prior and Outlier Suppression. . . . . . 138
Yiwei Zhang, Ge Li, Xiaoqiang Guo, Wenmin Wang,
and Ronggang Wang

Single Image Super-Resolution Using Multi-scale Convolutional

Neural Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
Xiaoyi Jia, Xiangmin Xu, Bolun Cai, and Kailing Guo

Person Identity and Emotion

A Novel Image Preprocessing Strategy for Foreground Extraction

in Person Re-identification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
Daiyin Wang, Wenbin Yao, and Yuesheng Zhu

Age Estimation via Pose-Invariant 3D Face Alignment Feature

in 3 Streams of CNN. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
Li Sun, Song Qiu, Qingli Li, Hongying Liu, and Mei Zhou

Face Alignment Using Local Probabilistic Features . . . . . . . . . . . . . . . . . . . 184

Qing Lu, Jun Yu, and Zengfu Wang

Multi-modal Emotion Recognition with Temporal-Band Attention

Based on LSTM-RNN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
Jiamin Liu, Yuanqi Su, and Yuehu Liu
 Contents – Part I XIX

Multimodal Fusion of Spatial-Temporal Features for Emotion

Recognition in the Wild. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
Zuchen Wang and Yuchun Fang

A Fast and General Method for Partial Face Recognition . . . . . . . . . . . . . . . 215

Qianhao Wu and Zechao Li

Tracking and Action Recognition

Adaptive Correlation Filter Tracking with Weighted Foreground

Representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
Chunguang Qie, Hanzi Wang, Yan Yan, Guanjun Guo,
and Jin Zheng

A Novel Method for Camera Pose Tracking Using Visual

Complementary Filtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
Xiangkai Lin and Ronggang Wang

Trajectory-Pooled 3D Convolutional Descriptors

for Action Recognition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
Xiusheng Lu, Hongxun Yao, Xiaoshuai Sun,
Shengping Zhang, and Yanhao Zhang

Temporal Interval Regression Network for Video Action Detection . . . . . . . . 258

Qing Wang, Laiyun Qing, Jun Miao, and Lijuan Duan

Semantic Sequence Analysis for Human Activity Prediction . . . . . . . . . . . . . 269

Guolong Wang, Zheng Qin, and Kaiping Xu

Motion State Detection Based Prediction Model for Body Parts

Tracking of Volleyball Players . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 280
Fanglu Xie, Xina Cheng, and Takeshi Ikenaga

Detection and Classification

Adapting Generic Detector for Semi-Supervised Pedestrian Detection . . . . . . 293

Shiyao Lei, Qiujia Ji, Shufeng Wang, and Si Wu

StairsNet: Mixed Multi-scale Network for Object Detection . . . . . . . . . . . . . 303

Weiyi Gao, Wenlong Cao, Jian Zhai, and Jianwu Rui

A Dual-CNN Model for Multi-label Classification by Leveraging

Co-occurrence Dependencies Between Labels . . . . . . . . . . . . . . . . . . . . . . . 315
Peng-Fei Zhang, Hao-Yi Wu, and Xin-Shun Xu

Multi-level Semantic Representation for Flower Classification . . . . . . . . . . . 325

Chuang Lin, Hongxun Yao, Wei Yu, and Wenbo Tang
XX Contents – Part I

Multi-view Multi-label Learning via Optimal Classifier Chain. . . . . . . . . . . . 336

Yiming Liu and Xingwei Hao

Tire X-ray Image Impurity Detection Based on Multiple

Kernel Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 346
Shuai Zhao, Zhineng Chen, Baokui Li, and Bin Zhang

Multimedia Signal Reconstruction and Recovery

CRF-Based Reconstruction from Narrow-Baseline Image Sequences . . . . . . . 359

Yue Xu, Qiuyan Tao, Lianghao Wang, Dongxiao Li, and Ming Zhang

Better and Faster, when ADMM Meets CNN: Compressive-Sensed

Image Reconstruction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 370
Chen Zhao, Ronggang Wang, and Wen Gao

Sparsity-Promoting Adaptive Coding with Robust Empirical Mode

Decomposition for Image Restoration . . . . . . . . . . . . . . . . . . . . . . . . . . . . 380
Rui Chen, Huizhu Jia, Xiaodong Xie, and Gao Wen

A Splicing Interpolation Method for Head-Related Transfer Function . . . . . . 390

Chunling Ai, Xiaochen Wang, Yafei Wu, and Cheng Yang

Structured Convolutional Compressed Sensing Based on Deterministic

Subsamplers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 400
Shu Wang, Zhongyuan Wang, and Yimin Luo

Blind Speech Deconvolution via Pretrained Polynomial Dictionary

and Sparse Representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 411
Jian Guan, Xuan Wang, Shuhan Qi, Jing Dong, and Wenwu Wang

Text and Line Detection/Recognition

Multi-lingual Scene Text Detection Based on Fully

Convolutional Networks. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 423
Shaohua Liu, Yan Shang, Jizhong Han, Xi Wang, Hongchao Gao,
and Dongqin Liu

Cloud of Line Distribution for Arbitrary Text Detection

in Scene/Video/License Plate Images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 433
Wenhai Wang, Yirui Wu, Shivakumara Palaiahnakote, Tong Lu,
and Jun Liu

Affine Collaborative Representation Based Classification for In-Air

Handwritten Chinese Character Recognition . . . . . . . . . . . . . . . . . . . . . . . . 444
Jianshe Zhou, Zhaochun Xu, Jie Liu, Weiqiang Wang, and Ke Lu
 Contents – Part I XXI

Overlaid Chinese Character Recognition via a Compact CNN. . . . . . . . . . . . 453

Hongzhu Li and Weiqiang Wang

Efficient and Robust Lane Detection Using Three-Stage Feature

Extraction with Line Fitting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 464
Aming Wu and Yahong Han

Social Media

Saliency-GD: A TF-IDF Analogy for Landmark Image Mining. . . . . . . . . . . 477

Wei Li, Jianmin Li, and Bo Zhang

An Improved Clothing Parsing Method Emphasizing the Clothing

with Complex Texture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 487
Juan Ji and Ruoyu Yang

Detection of Similar Geo-Regions Based on Visual Concepts

in Social Photos . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 497
Hiroki Takimoto, Magali Philippe, Yasutomo Kawanishi,
Ichiro Ide, Takatsugu Hirayama, Keisuke Doman,
Daisuke Deguchi, and Hiroshi Murase

Unsupervised Concept Learning in Text Subspace for Cross-Media

Retrieval . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 505
Mengdi Fan, Wenmin Wang, Peilei Dong, Ronggang Wang,
and Ge Li

Image Stylization for Thread Art via Color Quantization

and Sparse Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 515
Kewei Yang, Zhengxing Sun, Shuang Wang,
and Hui-Hsia Chen

Least-Squares Regulation Based Graph Embedding . . . . . . . . . . . . . . . . . . . 526

Si-Xing Liu, Timothy Apasiba Abeo, and Xiang-Jun Shen

SSGAN: Secure Steganography Based on Generative

Adversarial Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 534
Haichao Shi, Jing Dong, Wei Wang, Yinlong Qian,
and Xiaoyu Zhang

Generating Chinese Poems from Images Based on Neural Network . . . . . . . . 545

Shuo Xing, Xueliang Liu, Richang Hong, and Ye Zhao

Detail-Enhancement for Dehazing Method Using Guided Image

Filter and Laplacian Pyramid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 555
Dong Zhao and Long Xu
XXII Contents – Part I

Personalized Micro-Video Recommendation via Hierarchical

User Interest Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 564
Lei Huang and Bin Luo

3D and Panoramic Vision

MCTD: Motion-Coordinate-Time Descriptor for 3D Skeleton-Based

Action Recognition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 577
Qi Liang and Feng Wang

Dense Frame-to-Model SLAM with an RGB-D Camera . . . . . . . . . . . . . . . . 588

Xiaodan Ye, Jianing Li, Lianghao Wang, Dongxiao Li,
and Ming Zhang

Parallax-Robust Hexahedral Panoramic Video Stitching . . . . . . . . . . . . . . . . 598

Sha Guo, Ronggang Wang, Xiubao Jiang, Zhenyu Wang,
and Wen Gao

Image Formation Analysis and Light Field Information

Reconstruction for Plenoptic Camera 2.0 . . . . . . . . . . . . . . . . . . . . . . . . . . 609
Li Liu, Xin Jin, and Qionghai Dai

Part Detection for 3D Shapes via Multi-view Rendering . . . . . . . . . . . . . . . 619

Youcheng Song, Zhengxing Sun, Mofei Song, and Yunjie Wu

Benchmarking Screen Content Image Quality Evaluation in Spatial

Psychovisual Modulation Display System. . . . . . . . . . . . . . . . . . . . . . . . . . 629
Yuanchun Chen, Guangtao Zhai, Ke Gu, Xinfeng Zhang, Weisi Lin,
and Jiantao Zhou

A Fast Sample Adaptive Offset Algorithm for H.265/HEVC . . . . . . . . . . . . 641

Yan Zhou and Zhenzhong Chen

Blind Quality Assessment for Screen Content Images

by Texture Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 652
Ning Lu and Guohui Li

Assessment of Visually Induced Motion Sickness in Immersive Videos . . . . . 662

Huiyu Duan, Guangtao Zhai, Xiongkuo Min, Yucheng Zhu, Wei Sun,
and Xiaokang Yang

Hybrid Kernel-Based Template Prediction and Intra Block Copy

for Light Field Image Coding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 673
Deyang Liu, Ping An, Ran Ma, Xinpeng Huang, and Liquan Shen

Asymmetric Representation for 3D Panoramic Video. . . . . . . . . . . . . . . . . . 683

Guisen Xu, Yueming Wang, Zhenyu Wang, and Ronggang Wang
 Contents – Part I XXIII

Deep Learning for Signal Processing and Understanding

Shallow and Deep Model Investigation for Distinguishing

Corn and Weeds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 693
Yu Xia, Hongxun Yao, Xiaoshuai Sun, and Yanhao Zhang

Representing Discrimination of Video by a Motion Map . . . . . . . . . . . . . . . 703

Wennan Yu, Yuchao Sun, Feiwu Yu, and Xinxiao Wu

Multi-scale Discriminative Patches for Fined-Grained

Visual Categorization. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 712
Wenbo Tang, Hongxun Yao, Xiaoshuai Sun, and Wei Yu

Chinese Characters Recognition from Screen-Rendered Images

Using Inception Deep Learning Architecture. . . . . . . . . . . . . . . . . . . . . . . . 722
Xin Xu, Jun Zhou, Hong Zhang, and Xiaowei Fu

Visual Tracking by Deep Discriminative Map. . . . . . . . . . . . . . . . . . . . . . . 733

Wenyi Tang, Bin Liu, and Nenghai Yu

Hand Gesture Recognition by Using 3DCNN and LSTM

with Adam Optimizer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 743
Siyu Jiang and Yimin Chen

Learning Temporal Context for Correlation Tracking

with Scale Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 754
Yuhao Cui, Haoqian Wang, Xingzheng Wang, and Yi Yang

Deep Combined Image Denoising with Cloud Images . . . . . . . . . . . . . . . . . 764

Sifeng Xia, Jiaying Liu, Wenhan Yang, Mading Li,
and Zongming Guo

Vehicle Verification Based on Deep Siamese Network

with Similarity Metric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 773
Qian Zhang, Mingtao Pei, Mei Chen, and Yunde Jia

Style Transfer with Content Preservation from Multiple Images . . . . . . . . . . 783

Dilin Liu, Wei Yu, and Hongxun Yao

Task-Specific Neural Networks for Pose Estimation in Person

Re-identification Task . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 792
Kai Lv, Hao Sheng, Yanwei Zheng, Zhang Xiong, Wei Li,
and Wei Ke

Mini Neural Networks for Effective and Efficient Mobile

Album Organization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 802
Lingling Fa, Lifei Zhang, Xiangbo Shu, Yan Song,
and Jinhui Tang
XXIV Contents – Part I

Sweeper: Design of the Augmented Path in Residual Networks . . . . . . . . . . 811

Kang Shi and Weiqiang Wang

Large-Scale Multimedia Affective Computing

Sketch Based Model-Like Standing Style Recommendation . . . . . . . . . . . . . 825

Ying Zheng, Hongxun Yao, and Dong Wang

Joint L1 L2 Regularisation for Blind Speech Deconvolution . . . . . . . . . . . 834

Jian Guan, Xuan Wang, Zongxia Xie, Shuhan Qi,
and Wenwu Wang

Multi-modal Emotion Recognition Based on Speech and Image . . . . . . . . . . 844

Yongqiang Li, Qi He, Yongping Zhao, and Hongxun Yao

Analysis of Psychological Behavior of Undergraduates . . . . . . . . . . . . . . . . 854

Chunchang Gao

Sensor-Enhanced Multimedia Systems

Compression Artifacts Reduction for Depth Map by Deep

Intensity Guidance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 863
Pingping Zhang, Xu Wang, Yun Zhang, Lin Ma, Jianmin Jiang,
and Sam Kwong

LiPS: Learning Social Relationships in Probe Space . . . . . . . . . . . . . . . . . . 873

Chaoxi Li, Chengwen Luo, Junliang Chen, Hande Hong,
Jianqiang Li, and Long Cheng

The Intelligent Monitoring for the Elderly Based on WiFi Signals. . . . . . . . . 883
Nan Bao, Chengyang Wu, Qiancheng Liang, Lisheng Xu, Guozhi Li,
Ziyu Qi, Wanyi Zhang, He Ma, and Yan Li

Sentiment Analysis for Social Sensor. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 893

Xiaoyu Zhu, Tian Gan, Xuemeng Song, and Zhumin Chen

Recovering Overlapping Partials for Monaural Perfect Harmonic Musical

Sound Separation Using Modified Common Amplitude Modulation . . . . . . . 903
Yukai Gong, Xiangbo Shu, and Jinhui Tang

Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 913

Content Analysis
 A Competitive Combat Strategy and Tactics
in RTS Games AI and StarCraft

Adil Khan1 ✉ , Kai Yang2, Yunsheng Fu3, Fang Lou3, Worku Jifara1,
( )

Feng Jiang1, and Liu Shaohui1

1
School of Computer Science and Technology, Harbin Institute of Technology,
Harbin 150001, People’s Republic of China
adil.adil25@yahoo.com, worku.jifara@gmail.com,
{fjiang,shliu}@hit.edu.cn
2
Chinese People’s Liberation Army Aviation School, Beijing, People’s Republic of China
yangkai4545@163.com
3
Institute of Computer Applications, China Academy of Engineering Physics, Mianyang,
People’s Republic of China
Fyun@yahoo.com, Lfang@163.com

Abstract. This paper presents a competitive combat strategy and tactics in RTS
Games AI. To put it simply, if a player is building up base, he is losing out on
creating an army and If he is building up his army, he is losing out on having a
strong base. The key to winning, in StarCraft or any other RTS game is to balance
strategy, tactics, macro and micro. To improve the game, one has to be able to
keep track of everything that’s going on over the entire map. And one must be
able to give orders quickly and efficiently so in this paper we propose a compet‐
itive battle strategy with the help of a plot and decision tree. We simulate the
strategy in MicroRTS developed in java EE by conducting a game-play between
human player and MicroRTS AI (Game AI), though our proposed strategy out-
performs the Game AI rarely as we did not account game playing-speed that
makes a huge difference in victory but at least we succeeded in introducing a
strategy that could well compete the Game AI and may defeat it but rarely.

Keywords: Combat strategy · Game AI · MicroRTS · Macro management

Real time strategy games · StarCraft

1 Introduction

Real time Strategy games are combat games in which a player try to dominate or destroy
the opponent as quick as possible by managing and controlling resources, tactics and
strategies [1]. Real time strategy is a sub-genre of strategy video games where the games
do not progress or play incrementally [2]. The term ‘Real Time Strategy’ was first intro‐
duced in a magazine called BYTE in 1982 and the first game of this genre was Dune II
developed by Westwood Studio and released by Virgin Interactive in 1992 followed by
their seminal command & Conquer in 1995 [3]. On March 31, 1998 StarCraft was
released for windows which was the first game in the StarCraft series based on science
fiction real time strategy game and since then it grew in popularity among public in the

© Springer International Publishing AG, part of Springer Nature 2018

B. Zeng et al. (Eds.): PCM 2017, LNCS 10736, pp. 3–12, 2018.
https://doi.org/10.1007/978-3-319-77383-4_1
4 A. Khan et al.

market [4]. Now this game is being played every day by millions of players or users and
dominating and surpassing all other RTS games of this genre [5]. A ‘strategy game’ is
a game in which the players intimidate and most often make independent strategic and
tactical decisions that have a high importance in determining the outcome of the game.
Almost all strategy games oblige internal decision tree style thinking, and typically very
high situational consciousness or awareness [6].
Multiplayer online battle arena (MOBA) is a sub-genre of real time strategy video
games in which the objective is to dominate and destroy the enemy’s main structure like
Bases, units, buildings and other resources [7]. Players typically have several abilities
and benefits that improve over the progression of a game and that contribute in
destroying the opponent team’s overall strategy. Defense of the Ancients or “DotA” is
one of the MOBA game. Players use commanding units known as heroes, and are
supported by allied heroes and AI-controlled combatants called ‘creeps’. Some of the
best real time strategy games are StarCraft II: Wings of Liberty, StarCraft II: Legacy of
the Void, Rome: Total War, Company of Heroes, Warcraft III: Reign of Chaos, Rise of
Nations etc. We studied other related papers in the field that were either found outdated
and specific to Some traditional strategic concepts or found with complicated and hybrid
tactical and strategic techniques only, that is why we thought to write a paper with a
unique and simple strategy or tactics pursuing updated information and literature, so this
paper aims to introduce a simple but an efficient and competitive strategy for RTS games
in the perspective of the current and recent past research and state of the arts in RTS
games and StarCraft. It is organized as follows. Section 2 explains related work on RTS
Games-AI and StarCraft. Section 3 presents a simple RTS game strategy with its deci‐
sion tree. Section 4 includes experiments and implementation of the introduced strategy
and Sect. 5 concludes the paper.

2 Related Work

Real time strategy games are fast paced combat games that were first introduced in
1990’s and since then grew in a great popularity. RTS games face many challenges in
AI research such as RTS games are played on large maps on which large numbers of
units move around under player control collecting resources, constructing buildings,
scouting, and attacking opponents with the goal of destroying all enemy buildings but
this renders traditional full width search infeasible [8]. Combat strategy or map explo‐
ration is a common discovery oriented activity that players perform in modern video
games. Exploration is a game mechanism that players need to master in order to collect
resources and advance in the game [9]. In every RTS game, time is the most valuable
resource before thinking about other resources. A player needs to keep himself aware
of the time that he loses. It suggests that time is an important factor in composing a
combat strategy always because if the opponent is faster so then it does not matter how
clever the player is or how carefully the player base defenses and army compositions
are [10], this also does not mean that speed trumps intelligence in RTS games but the
speed is the foundation and backbone of an effective strategy. Along with speed, an
effective combat strategy must need to be well composed of balanced, smart planning
 A Competitive Combat Strategy and Tactics in RTS Games AI and StarCraft 5

and responsive tactics [11]. According to the research community in RTS games, speed
is more worthwhile than accuracy because in RTS games a player too often cannot afford
to spend enough time in deciding where to place or manage a building, base or any other
resource [12]. Developing computer controlled clusters to engage in fight, manage the
usage of restricted resources, and form units and buildings in real time strategy games
is a unique presentation in game-AI. But, tightly controlled online commercial game
carry challenges to researchers interested in observing player’s behavior and activities,
constructing player’s strategy models, and developing practical AI technology in them
[13]. In place of setting up new programming settings or building a huge extent of agent’s
decision instructions by player’s experience for directing real time AI research, the
authors used reruns or replays of the commercial RTS game StarCraft to assess human
player activities and to develop an intelligent system to learn humanoid decisions and
performances. A case based reasoning approach was applied for the objective of training
the system to learn and guess player strategies [14]. The researcher’s analysis demon‐
strated that the suggested system is capable of learning and expecting individual player
strategies, and that players provide signs of their private features through their building
production orders.

3 Proposed Strategy and Decision Tree

In this section we present a competitive combat game strategy followed by its Decision
tree with details as follows.

3.1 A Combat Strategy

Though it is not easy to program a game while planning any strategy for the first time
without those who know playing RTS games like StarCraft. We thought over different
strategies starting from rushes to complex micromanagement strategies. After all a
simple Macro strategy was decided which is to out resource the enemy (opponent) by
collecting the resources quickly than them that may result in victory in a long run. During
the design, obviously we had a focus on worker units because of their certain role in
success. During designing the units are required to act according to the strategy in finding
and taking the remote resources to the base, if there is not any base near then it would
start building one. Attacking units normally find the target and attack but sometimes
workers will defend if they needed to, besides the bases and barrack will also keep
producing attackers and workers to combat. We thought of a strong defense supporting
the rush strategy by introducing the defense team that will guard the bases and will wait
targeting until someone comes close. A percentage of defenders are assigned to attacking
units called ‘Assaults’ to attack enemy when they are required or when they exceed a
specific size or number. During units’ production the ratio of defenders and assault is
always kept balance. Also with the passage of time new units are assigned to assault
team as the force grows larger. At least one of the three attack units i.e. Light, heavy
and ranged would be opted to build at a specific time. For the proposed strategy we
decided to work with a decision Tree, being one of the simplest, easiest to implement,
Another random document with
no related content on Scribd:
 July 2.—Almost the Glorious Fourth of July. How
shall we celebrate? Know of no way except to pound
on the bake tin, which I shall do. Have taken to
rubbing my limbs, which are gradually becoming
more dropsical. Badly swollen. One of my teeth came
out a few days ago, and all are loose. Mouth very
sore. Battese says: “We get away yet.” Works around
and always busy. If any news, he merely listens and
don’t say a word. Even he is in poor health, but never
mentions it. An acquaintance of his says he owns a
good farm in Minnesota. Asked him if he was married
—says: “Oh, yes.” Any children? “Oh, yes.” This is as
far as we have got his history. Is very different from
Indians in general. Some of them here are
despisable cowards—worse than the negro.
Probably one hundred negroes are here. Not so
tough as the whites. Dead line being fixed up by the
rebels. Got down in some places. Bought a piece of
soap, first I have seen in many months. Swamp now
in frightful condition from the filth of camp. Vermin
and raiders have the best of it. Capt. Moseby still
leads the villains.
 THE RAIDERS PUT DOWN.

ANDERSONVILLE ON ITS METAL—LEADING RAIDERS

ARRESTED, TRIED AND HUNG—GREAT EXCITEMENT FOR A
FEW DAYS, FOLLOWED BY GOOD ORDER—DEATH RATE
INCREASES, HOWEVER—THE ASTOR HOUSE MESS AS
POLICEMEN.

July 3.—Three hundred and fifty new men from

West Virginia were turned into this summer resort
this morning. They brought good news as to
successful termination of the war, and they also
caused war after coming among us. As usual the
raiders proceeded to rob them of their valuables and
a fight occurred in which hundreds were engaged.
The cut-throats came out ahead. Complaints were
made to Capt. Wirtz that this thing would be tolerated
no longer, that these raiders must be put down or the
men would rise in their might and break away if
assistance was not given with which to preserve
order. Wirtz flew around as if he had never thought of
it before, issued an order to the effect that no more
food would be given us until the leaders were
arrested and taken outside for trial. The greatest
possible excitement. Hundreds that have before
been neutral and non-commital are now joining a
police force. Captains are appointed to take charge
of the squads which have been furnished with clubs
by Wirtz. As I write, this middle of the afternoon, the
battle rages. The police go right to raider head-
quarters knock right and left and make their arrests.
Sometimes the police are whipped and have to
retreat, but they rally their forces and again make a
charge in which they are successful. Can lay in our
shade and see the trouble go on. Must be killing
some by the shouting. The raiders fight for their very
life, and are only taken after being thoroughly
whipped. The stockade is loaded with guards who
are fearful of a break. I wish I could describe the
scene to-day. A number killed. After each arrest a
great cheering takes place. Night.—Thirty or forty
have been taken outside of the worst characters in
camp, and still the good work goes on. No food to-
day and don’t want any. A big strapping fellow called
Limber Jim heads the police. Grand old Michael
Hoare is at the front and goes for a raider as quick as
he would a rebel. Patrol the camp all the time and
gradually quieting down. The orderly prisoners are
feeling jolly.
July 4.—The men taken outside yesterday are
under rebel guard and will be punished. The men are
thoroughly aroused, and now that the matter has
been taken in hand, it will be followed up to the letter.
Other arrests are being made to-day, and
occasionally a big fight. Little Terry, whom they could
not find yesterday, was to-day taken. Had been
hiding in an old well, or hole in the ground. Fought
like a little tiger, but had to go. “Limber Jim” is a brick,
and should be made a Major General if he ever
reaches our lines. Mike Hoare is right up in rank, and
true blue. Wm. B. Rowe also makes a good
policeman, as does “Dad” Sanders. Battese says he
“no time to fight, must wash.” Jimmy Devers regrets
that he cannot take a hand in, as he likes to fight,
and especially with a club. The writer hereof does no
fighting, being on the sick list. The excitement of
looking on is most too much for me. Can hardly
arrest the big graybacks crawling around. Capt.
Moseby is one of the arrested ones. His right name is
Collins and he has been in our hundred all the time
since leaving Richmond. Has got a good long neck to
stretch. Another man whom I have seen a good deal
of, one Curtiss, is also arrested. I haven’t mentioned
poor little Bullock for months, seems to me. He was
most dead when we first came to Andersonville, and
is still alive and tottering around. Has lost his voice
entirely and is nothing but a skeleton. Hardly enough
of him for disease to get hold of. Would be one of the
surprising things on record if he lives through it, and
he seems no worse than months ago. It is said that a
court will be formed of our own men to try the raiders.
Any way, so they are punished. All have killed men,
and they themselves should be killed. When
arrested, the police had hard work to prevent their
being lynched. Police more thoroughly organizing all
the time. An extra amount of food this p. m., and
police get extra rations, and three out of our mess is
doing pretty well, as they are all willing to divide.
They tell us all the encounters they have, and much
interesting talk. Mike has some queer experiences.
Rebel flags at half mast for some of their great men.
Just heard that the trial of raiders will begin to-
morrow.
July 5.—Court is in session outside and raiders
being tried by our own men. Wirtz has done one
good thing, but it’s a question whether he is entitled
to any credit, as he had to be threatened with a break
before he would assist us. Rations again to-day. I am
quite bad off with my diseases, but still there are so
many thousands so much worse off that I do not
complain much, or try not to however.
July 6.—Boiling hot, camp reeking with filth, and no
sanitary privileges; men dying off over a hundred and
forty per day. Stockade enlarged, taking in eight or
ten more acres, giving us more room, and stumps to
dig up for wood to cook with. Mike Hoare is in good
health; not so Jimmy Devers. Jimmy has now been a
prisoner over a year, and poor boy, will probably die
soon. Have more mementoes than I can carry, from
those who have died, to be given to their friends at
home. At least a dozen have given me letters,
pictures &c., to take North. Hope I shan’t have to turn
them over to some one else.
July 7.—The court was gotten up by our own men
and from our own men; Judge, jury, counsel, &c. Had
a fair trial, and were even defended, but to no
purpose. It is reported that six have been sentenced
to be hung, while a good many others are
condemned to lighter punishment, such as setting in
the stocks, strung up by the thumbs, thumb screws,
head hanging, etc. The court has been severe, but
just. Mike goes out to-morrow to take some part in
the court proceedings. The prison seems a different
place altogether; still, dread disease is here, and
mowing down good and true men. Would seem to
me that three or four hundred died each day, though
officially but one hundred and forty odd is told. About
twenty-seven thousand, I believe, are here now in all.
No new ones for a few days. Rebel visitors, who look
at us from a distance. It is said the stench keeps all
away who have no business here and can keep
away. Washing business good. Am negotiating for a
pair of pants. Dislike fearfully to wear dead men’s
clothes, and haven’t to any great extent.
July 8.—Oh, how hot, and oh, how miserable. The
news that six have been sentenced to be hanged is
true, and one of them is Moseby. The camp is
thoroughly under control of the police now, and it is a
heavenly boon. Of course there is some stealing and
robbery, but not as before. Swan, of our mess, is sick
with scurvy. I am gradually swelling up and growing
weaker. But a few more pages in my diary. Over a
hundred and fifty dying per day now, and twenty-six
thousand in camp. Guards shoot now very often.
Boys, as guards, are the most cruel. It is said that if
they kill a Yankee, they are given a thirty days
furlough. Guess they need them as soldiers too
much to allow of this. The swamp now is fearful,
water perfectly reeking with prison offal and poison.
Still men drink it and die. Rumors that the six will be
hung inside. Bread to-day and it is so coarse as to do
more hurt than good to a majority of the prisoners.
The place still gets worse. Tunneling is over with; no
one engages in it now that I know of. The prison is a
success as regards safety; no escape except by
death, and very many take advantage of that way. A
man who has preached to us (or tried to) is dead.
Was a good man I verily believe, and from
Pennsylvania. It’s almost impossible for me to get
correct names to note down; the last named man
was called “the preacher,” and I can find no other
name for him. Our quartette of singers a few rods
away is disbanded. One died, one nearly dead, one a
policeman and the other cannot sing alone, and so
where we used to hear and enjoy good music
evenings, there is nothing to attract us from the
groans of the dying. Having formed a habit of going
to sleep as soon as the air got cooled off and before
fairly dark, I wake up at two or three o’clock and stay
awake. I then take in all the horrors of the situation.
Thousands are groaning, moaning and crying, with
no bustle of the day time to drown it. Guards every
half hour call out the time and post, and there is often
a shot to make one shiver as if with the ague. Must
arrange my sleeping hours to miss getting owly in the
morning. Have taken to building air castles of late, on
being exchanged. Getting loony, I guess, same as all
the rest.
 Michael Hoare.
Now an inmate of the Home for Disabled
Soldiers, Dayton, Ohio.

July 9.—Battese brought me some onions, and if

they ain’t good then no matter; also a sweet potato.
One-half the men here would get well if they only had
something in the vegetable line to eat, or acids.
Scurvy is about the most loathsome disease, and
when dropsy takes hold with the scurvy, it is terrible. I
have both diseases but keep them in check, and it
only grows worse slowly. My legs are swollen, but the
cords are not contracted much, and I can still walk
very well. Our mess all keep clean, in fact are obliged
to or else turned adrift. We want none of the dirty sort
in our mess. Sanders and Rowe enforce the rules,
which is not much work, as all hands are composed
of men who prefer to keep clean. I still do a little
washing, but more particularly hair cutting, which is
easier work. You should see one of my hair cuts.
Nobby! Old prisoners have hair a foot long or more,
and my business is to cut it off, which I do without
regards to anything except to get it off. I should judge
that there are one thousand rebel soldiers guarding
us, and perhaps a few more, with the usual number
of officers. A guard told me to-day that the yanks
were “gittin licked,” and they didn’t want us
exchanged; just as soon we should die here as not; a
yank asked him if he knew what exchange meant;
said he knew what shootin’ meant, and as he began
to swing around his old shooting iron we retreated in
among the crowd. Heard that there were some new
men belonging to my regiment in another part of the
prison; have just returned from looking after them
and am all tired out. Instead of belonging to the 9th
Michigan Cavalry, they belong to the 9th Michigan
Infantry. Had a good visit and quite cheered with their
accounts of the war news. Some one stole Battese’s
wash board and he is mad; is looking for it—may
bust up the business. Think Hub Dakin will give me a
board to make another one. Sanders owns the jack
knife, of this mess, and he don’t like to lend it either;
borrow it to carve on roots for pipes. Actually take
solid comfort “building castles in the air,” a thing I
have never been addicted to before. Better than
getting blue and worrying myself to death. After all,
we may get out of this dod-rotted hole. Always an
end of some sort to such things.
July 10.—Have bought of a new prisoner quite a
large (thick I mean,) blank book so as to continue my
diary. Although it’s a tedious and tiresome task, am
determined to keep it up. Don’t know of another man
in prison who is doing likewise. Wish I had the gift of
description that I might describe this place. Know that
I am not good at such things, and have more
particularly kept track of the mess which was the
“Astor House Mess” on Belle Isle, and is still called
so here. Thought that Belle Isle was a very bad
place, and used about the worst language I knew
how to use in describing it, and so find myself at fault
in depicting matters here as they are. At Belle Isle we
had good water and plenty of it, and I believe it
depends more upon water than food as regards
health. We also had good pure air from up the James
River. Here we have the very worst kind of water.
Nothing can be worse or nastier than the stream
drizzling its way through this camp. And for air to
breathe, it is what arises from this foul place. On all
four sides of us are high walls and tall trees, and
there is apparently no wind or breeze to blow away
the stench, and we are obliged to breathe and live in
it. Dead bodies lay around all day in the broiling sun,
by the dozen and even hundreds, and we must suffer
and live in this atmosphere. It’s too horrible for me to
describe in fitting language. There was once a very
profane man driving a team of horses attached to a
wagon in which there were forty or fifty bushels of
potatoes. It was a big load and there was a long hill
to go up. The very profane man got off the load of
potatoes to lighten the weight, and started the team
up the hill. It was hard work, but they finally reached
the top and stopped to rest. The profane man looked
behind him and saw that the end board of the wagon
had slipped out just as he had started, and there the
potatoes were, scattered all the way along up the hill.
Did the man make the very air blue with profanity?
No, he sat down on a log feeling that he couldn’t do
the subject justice and so he remarked: “No! it’s no
use, I can’t do it justice.” While I have no reason or
desire to swear, I certainly cannot do this prison
justice. It’s too stupenduous an undertaking. Only
those who are here will ever know what
Andersonville is.

AN ACCOUNT OF THE HANGING.

July 11.—This morning lumber was brought into
the prison by the rebels, and near the gate a gallows
erected for the purpose of executing the six
condemned Yankees. At about ten o’clock they were
brought inside by Capt. Wirtz and some guards, and
delivered over to the police force. Capt. Wirtz then
said a few words about their having been tried by our
own men and for us to do as we choose with them,
that he washed his hands of the whole matter, or
words to that effect. I could not catch the exact
language, being some little distance away. I have
learned by enquiry, their names, which are as
follows: John Sarsfield, 144th New York; William
Collins, alias “Moseby,” Co. D, 88th Pennsylvania;
Charles Curtiss, Battery A, 5th Rhode Island Artillery;
Pat Delaney, Co. E, 83d Pennsylvania; A. Munn, U.
S. Navy, and W. R. Rickson of the U. S. Navy. After
Wirtz made his speech he withdrew his guards,
leaving the condemned at the mercy of 28,000
enraged prisoners who had all been more or less
wronged by these men. Their hands were tied behind
them, and one by one they mounted the scaffold.
Curtiss, who was last, a big stout fellow, managed to
get his hands loose and broke away and ran through
the crowd and down toward the swamp. It was yelled
out that he had a knife in his hand, and so a path
was made for him. He reached the swamp and
plunged in, trying to get over on the other side,
presumably among his friends. It being very warm he
over exerted himself, and when in the middle or
thereabouts, collapsed and could go no farther. The
police started after him, waded in and helped him
out. He pleaded for water and it was given him. Then
led back to the scaffold and helped to mount up. All
were given a chance to talk. Munn, a good looking
fellow in marine dress, said he came into the prison
four months before perfectly honest, and as innocent
of crime as any fellow in it. Starvation, with evil
companions, had made him what he was. He spoke
of his mother and sisters in New York, that he cared
nothing as far as he himself was concerned, but the
news that would be carried home to his people made
him want to curse God he had ever been born.
Delaney said he would rather be hung than live here
as the most of them lived, on their allowance of
rations. If allowed to steal could get enough to eat,
but as that was stopped had rather hang. Bid all
good bye. Said his name was not Delaney and that
no one knew who he really was, therefore his friends
would never know his fate, his Andersonville history
dying with him. Curtiss said he didn’t care a ——,
only hurry up and not be talking about it all day;
making too much fuss over a very small matter.
William Collins, alias Moseby, said he was innocent
of murder and ought not to be hung; he had stolen
blankets and rations to preserve his own life, and
begged the crowd not to see him hung as he had a
wife and child at home, and for their sake to let him
live. The excited crowd began to be impatient for the
“show” to commence as they termed it. Sarsfield
made quite a speech; he had studied for a lawyer; at
the outbreak of the rebellion he had enlisted and
served three years in the army, been wounded in
battle, furloughed home, wound healed up, promoted
to first sergeant and also commissioned; his
commission as a lieutenant had arrived but had not
been mustered in when he was taken prisoner;
began by stealing parts of rations, gradually
becoming hardened as he became familiar with the
crimes practiced; evil associates had helped him to
go down hill and here he was. The other did not care
to say anything. While the men were talking they
were interrupted by all kinds of questions and
charges made by the crowd, such as “don’t lay it on
too thick, you villain,” “get ready to jump off,” “cut it
short,” “you was the cause of so and so’s death,”
“less talk and more hanging,” &c., &c. At about
eleven o’clock they were all blindfolded, hands and
feet tied, told to get ready, nooses adjusted and the
plank knocked from under. Moseby’s rope broke and
he fell to the ground, with blood spurting from his
ears, mouth and nose. As they was lifting him back to
the swinging off place he revived and begged for his
life, but no use, was soon dangling with the rest, and
died very hard. Munn died easily, as also did
Delaney, all the rest died hard and particularly
Sarsfield who drew his knees nearly to his chin and
then straightened them out with a jerk, the veins in
his neck swelling out as if they would burst. It was an
awful sight to see, still a necessity. Moseby, although
he said he had never killed any one, and I don’t
believe he ever did deliberately kill a man, such as
stabbing or pounding a victim to death, yet he has
walked up to a poor sick prisoner on a cold night and
robbed him of blanket, or perhaps his rations and if
necessary using all the force necessary to do it.
These things were the same as life to the sick man,
for he would invariably die. The result has been that
many have died from his robbing propensities. It was
right that he should hang, and he did hang most
beautifully and Andersonville is the better off for it.
None of the rest denied that they had killed men, and
probably some had murdered dozens. It has been a
good lesson; there are still bad ones in camp but we
have the strong arm of the law to keep them in
check. All during the hanging scene the stockade
was covered with rebels, who were fearful a break
would be made if the raiders should try and rescue
them. Many citizens too were congregated on the
outside in favorable positions for seeing. Artillery was
pointed at us from all directions ready to blow us all
into eternity in short order; Wirtz stood on a high
platform in plain sight of the execution and says we
are a hard crowd to kill our own men. After hanging
for half an hour or so the six bodies were taken down
and carried outside. In noting down the speeches
made by the condemned men, have used my own
language; in substance it is the same as told by
them. I occupied a near position to the hanging and
saw it all from first to last, and stood there until they
were taken down and carried away. Was a strange
sight to see and the first hanging I ever witnessed.
The raiders had many friends who crowded around
and denounced the whole affair and but for the police
there would have been a riot; many both for and
against the execution were knocked down. Some will
talk and get into trouble thereby; as long as it does
no good there is no use in loud talk and exciting
arguments; is dangerous to advance any argument,
men are so ready to quarrel. Have got back to my
quarters thoroughly prostrated and worn out with
fatigue and excitement, and only hope that to-day’s
lesson will right matters as regards raiding. Battese
suspended washing long enough to look on and see
them hang and grunted his approval. Have omitted to
say that the good Catholic priest attended the
condemned. Rebel negroes came inside and began
to take down the scaffold; prisoners took hold to help
them and resulted in its all being carried off to
different parts of the prison to be used for kindling
wood, and the rebels get none of it back and are
mad. The ropes even have been gobbled up, and I
suppose sometime may be exhibited at the north as
mementoes of to-day’s proceedings. Mike Hoare
assisted at the hanging. Some fears are entertained
that those who officiated will get killed by the friends
of those hanged. The person who manipulated the
“drop,” has been taken outside on parole of honor, as
his life would be in danger in here. Jimmy thanks
God that he has lived to see justice done the raiders;
he is about gone—nothing but skin and bone and
can hardly move hand or foot; rest of the mess
moderately well. The extra rations derived from our
three mess-mates as policemen, helps wonderfully to
prolong life. Once in a while some of them gets a
chance to go outside on some duty and buy onions
or sweet potatoes which is a great luxury.
July 12.—Good order has prevailed since the
hanging. The men have settled right down to the
business of living, with no interruption. I keep thinking
our situation can get no worse, but it does get worse
every day and not less than one hundred and sixty
die each twenty-four hours. Probably one-fourth or
one-third of these die inside the stockade, the
balance in the hospital outside. All day and up to four
o’clock p. m., the dead are being gathered up and
carried to the south gate and placed in a row inside
the dead line. As the bodies are stripped of their
clothing in most cases as soon as the breath leaves,
and in some cases before, the row of dead presents
a sickening appearance. Legs drawn up and in all
shapes. They are black from pitch pine smoke and
laying in the sun. Some of them lay there for twenty
hours or more, and by that time are in a horrible
condition. At four o’clock a four or six mule wagon
comes up to the gate and twenty or thirty bodies are
loaded on to the wagon and they are carted off to be
put in trenches, one hundred in each trench, in the
cemetery, which is eighty or a hundred rods away.
There must necessarily be a great many whose
names are not taken. It is the orders to attach the
name, company and regiment to each body, but it is
not always done. I was invited to-day to dig in a
tunnel, but had to decline. My digging days are over.
Must dig now to keep out of the ground, I guess. It is
with difficulty now that I can walk, and only with the
help of two canes.
July 13.—Can see in the distance the cars go
poking along by this station, with wheezing old
engines, snorting along. As soon as night comes a
great many are blind, caused by sleeping in the open
air, with moon shining in the face. Many holes are
dug and excavations made in camp. Near our
quarters is a well about five or six feet deep, and the
poor blind fellows fall into this pit hole. None
seriously hurt, but must be quite shaken up. Half of
the prisoners have no settled place for sleeping,
wander and lay down wherever they can find room.
Have two small gold rings on my finger, worn ever
since I left home. Have also a small photograph
album with eight photographs in. Relics of civilization.
Should I get these things through to our lines they
will have quite a history. When I am among the
rebels I wind a rag around my finger to cover up the
rings, or else take them and put in my pocket. Bad off
as I have been, have never seen the time yet that I
would part with them. Were presents to me, and the
photographs have looked at about one-fourth of the
time since imprisonment. One prisoner made some