You are on page 1of 54

Computer Vision and Graphics

International Conference ICCVG 2018


Warsaw Poland September 17 19 2018
Proceedings Leszek J. Chmielewski
Visit to download the full and correct content document:
https://textbookfull.com/product/computer-vision-and-graphics-international-conferenc
e-iccvg-2018-warsaw-poland-september-17-19-2018-proceedings-leszek-j-chmielews
ki/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...

Computer Vision and Graphics International Conference


ICCVG 2020 Warsaw Poland September 14 16 2020
Proceedings Leszek J. Chmielewski

https://textbookfull.com/product/computer-vision-and-graphics-
international-conference-iccvg-2020-warsaw-poland-
september-14-16-2020-proceedings-leszek-j-chmielewski/

Computer Vision ECCV 2018 15th European Conference


Munich Germany September 8 14 2018 Proceedings Part III
Vittorio Ferrari

https://textbookfull.com/product/computer-vision-eccv-2018-15th-
european-conference-munich-germany-
september-8-14-2018-proceedings-part-iii-vittorio-ferrari/

Computer Vision ECCV 2018 15th European Conference


Munich Germany September 8 14 2018 Proceedings Part IV
Vittorio Ferrari

https://textbookfull.com/product/computer-vision-eccv-2018-15th-
european-conference-munich-germany-
september-8-14-2018-proceedings-part-iv-vittorio-ferrari/

Computer Vision ECCV 2018 15th European Conference


Munich Germany September 8 14 2018 Proceedings Part VII
Vittorio Ferrari

https://textbookfull.com/product/computer-vision-eccv-2018-15th-
european-conference-munich-germany-
september-8-14-2018-proceedings-part-vii-vittorio-ferrari/
Computer Vision ECCV 2018 15th European Conference
Munich Germany September 8 14 2018 Proceedings Part I
Vittorio Ferrari

https://textbookfull.com/product/computer-vision-eccv-2018-15th-
european-conference-munich-germany-
september-8-14-2018-proceedings-part-i-vittorio-ferrari/

Computer Vision ECCV 2018 15th European Conference


Munich Germany September 8 14 2018 Proceedings Part X
Vittorio Ferrari

https://textbookfull.com/product/computer-vision-eccv-2018-15th-
european-conference-munich-germany-
september-8-14-2018-proceedings-part-x-vittorio-ferrari/

Computer Vision ECCV 2018 15th European Conference


Munich Germany September 8 14 2018 Proceedings Part
XIII Vittorio Ferrari

https://textbookfull.com/product/computer-vision-eccv-2018-15th-
european-conference-munich-germany-
september-8-14-2018-proceedings-part-xiii-vittorio-ferrari/

Computer Vision ECCV 2018 15th European Conference


Munich Germany September 8 14 2018 Proceedings Part XIV
Vittorio Ferrari

https://textbookfull.com/product/computer-vision-eccv-2018-15th-
european-conference-munich-germany-
september-8-14-2018-proceedings-part-xiv-vittorio-ferrari/

Information Systems Security: 14th International


Conference, ICISS 2018, Bangalore, India, December
17-19, 2018, Proceedings Vinod Ganapathy

https://textbookfull.com/product/information-systems-
security-14th-international-conference-iciss-2018-bangalore-
india-december-17-19-2018-proceedings-vinod-ganapathy/
Leszek J. Chmielewski
Ryszard Kozera
Arkadiusz Orłowski
Konrad Wojciechowski
Alfred M. Bruckstein
Nicolai Petkov (Eds.)
LNCS 11114

Computer Vision
and Graphics
International Conference, ICCVG 2018
Warsaw, Poland, September 17–19, 2018
Proceedings

123
Lecture Notes in Computer Science 11114
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/7412
Leszek J. Chmielewski Ryszard Kozera

Arkadiusz Orłowski Konrad Wojciechowski


Alfred M. Bruckstein Nicolai Petkov (Eds.)


Computer Vision
and Graphics
International Conference, ICCVG 2018
Warsaw, Poland, September 17–19, 2018
Proceedings

123
Editors
Leszek J. Chmielewski Konrad Wojciechowski
Faculty of Applied Informatics Institute of Computer Science
and Mathematics Silesian University of Technology
Warsaw University of Life Sciences Gliwice, Poland
Warsaw, Poland
and
Ryszard Kozera
Faculty of Applied Informatics Polish-Japanese Academy of Information
and Mathematics Technology
Warsaw University of Life Sciences Warsaw, Poland
Warsaw, Poland
Alfred M. Bruckstein
Technion, Israel Institute of Technology
Arkadiusz Orłowski Haifa, Israel
Faculty of Applied Informatics
and Mathematics Nicolai Petkov
Warsaw University of Life Sciences University of Groningen
Warsaw, Poland Groningen, The Netherlands

ISSN 0302-9743 ISSN 1611-3349 (electronic)


Lecture Notes in Computer Science
ISBN 978-3-030-00691-4 ISBN 978-3-030-00692-1 (eBook)
https://doi.org/10.1007/978-3-030-00692-1

Library of Congress Control Number: 2018954669

LNCS Sublibrary: SL6 – Image Processing, Computer Vision, Pattern Recognition, and Graphics

© Springer Nature Switzerland AG 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.

This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface

The International Conference on Computer Vision and Graphics, organized since 2002,
is the continuation of The International Conferences on Computer Graphics and Image
Processing, GKPO, held in Poland every second year from 1990 to 2000. The founder
and organizer of these conferences was Prof. Wojciech Mokrzycki. The main objective
of ICCVG is to provide an environment for the exchange of ideas between researchers
in the closely related domains of computer vision and computer graphics.
ICCVG 2018 brought together 108 authors. The proceedings contain 45 papers,
each accepted on the grounds of merit and relevance confirmed by three independent
reviewers. The number of papers is smaller than in the previous years because we have
greatly increased the quality requirements in the reviewing process.
ICCVG 2018 was organized by the Association for Image Processing, Poland
(Towarzystwo Przetwarzania Obrazów – TPO), the Faculty of Applied Informatics and
Mathematics, Warsaw University of Life Sciences (WZIM SGGW), together with the
Faculty of Information Science, West Pomeranian University of Technology (WI
ZUT), Szczecin, and the Polish-Japanese Academy of Information Technology
(PJATK) as the supporting organizers.
The Association for Image Processing integrates the Polish community working on
the theory and applications of computer vision and graphics. It was formed between
1989 and 1991.
The Faculty of Applied Informatics and Mathematics (WZIM), established in 2008
at the Warsaw University of Life Sciences (SGGW), which celebrates its 10th
anniversary this year, offers programs of study in Informatics as well as in Informatics
and Econometrics. Its location at the leading life sciences university in Poland is the
source of opportunities for valuable research at the border of applied information
sciences, agribusiness, forestry, furniture and wood industry, veterinary medicine, and
the broadly understood domains of biology and economy.
We would like to thank all the members of the Scientific Committee, as well as the
additional reviewers, for their help in ensuring the high quality of the papers. We would
also like to thank Grażyna Domańska-Żurek for her excellent work on technically
editing the proceedings, and Dariusz Frejlichowski, Bartosz Świderski, Henryk Palus,
Grzegorz Gawdzik, Halina Paluszkiewicz-Schaitter, Dominika Rudaś, Beata Sztab, and
Aneta Ryńska for their engagement in the conference organization and administration.

September 2018 Leszek J. Chmielewski


Ryszard Kozera
Arkadiusz Orłowski
Konrad Wojciechowski
Alfred M. Bruckstein
Nicolai Petkov
Organization

• Association for Image Processing (TPO)


• Faculty of Applied Informatics and Mathematics,
Warsaw University of Life Sciences (WZIM SGGW)
• Polish-Japanese Academy of Information Technology (PJATK)
• Faculty of Computer Science and Information Technology,
West Pomeranian University of Technology (WI ZUT)
• Springer, Lecture Notes in Computer Science (LNCS)

Conference General Chairs

Leszek J. Chmielewski, Poland


Ryszard Kozera, Poland
Arkadiusz Orłowski, Poland
Konrad Wojciechowski, Poland

Scientific Committee

Ivan Bajla, Slovakia Adam Jóźwik, Poland


Gunilla Borgefors, Sweden Heikki Kälviäinen, Finland
Nadia Brancati, Italy Andrzej Kasiński, Poland
M. Emre Celebi, USA Włodzimierz Kasprzak, Poland
Leszek Chmielewski, Poland Bertrand Kerautret, France
Dmitry Chetverikov, Hungary Nahum Kiryati, Israel
Piotr Czapiewski, Poland Reinhard Klette, New Zealand
László Czúni, Hungary Przemysław Klęsk, Poland
Silvana Dellepiane, Italy Józef Korbicz, Poland
Marek Domański, Poland Marcin Korzeń, Poland
Mariusz Flasiński, Poland Ryszard Kozera, Poland
Paweł Forczmański, Poland Hans-Jörg Kreowski, Germany
Dariusz Frejlichowski, Poland Adam Krzyżak, Canada
Maria Frucci, Italy Juliusz L. Kulikowski, Poland
André Gagalowicz, France Marek Kurzyński, Poland
Duncan Gillies, UK Bogdan Kwolek, Poland
Samuel Morillas Gómez, Spain Y. B. Kwon, South Korea
Ewa Grabska, Poland Bart Lamiroy, France
Diego Gragnaniello, Italy Piotr Lech, Poland
Marcin Iwanowski, Poland Anna Lewandowska, Poland
VIII Organization

Dongwei Liu, New Zealand Volodymyr Ponomaryov, Mexico


Vladimir Lukin, Russia Piotr Porwik, Poland
Wojciech Maleika, Poland Edward Półrolniczak, Poland
Witold Malina, Poland Artur Przelaskowski, Poland
Krzysztof Małecki, Poland Giuliana Ramella, Italy
Radosşaw Mantiuk, Poland Ferran Reverter, Spain
Tomasz Marciniak, Poland Przemysław Rokita, Poland
Andrzej Materka, Poland Khalid Saeed, Poland
Nikolaos Mavridis, UAE Bok-Suk Shin, New Zealand
Przemysław Mazurek, Poland Samuel Silva, Portugal
Tomasz Mąka, Poland Gerald Schaefer, UK
Wojciech Mokrzycki, Poland Andrzej Śluzek, UAE
Mariusz Nieniewski, Poland Maciej Smiatacz, Poland
Sławomir Nikiel, Poland Bogdan Smołka, Poland
Lyle Noakes, Australia Ryszard Tadeusiewicz, Poland
Antoni Nowakowski, Poland João Manuel R. S. Tavares, Portugal
Adam Nowosielski, Poland Hendrik Thamer, Germany
Krzysztof Okarma, France Ventzeslav Valev, USA
Maciej Orkisz, France Libor Vasa, Czech Republic
Arkadiusz Orłowski, Poland Konrad Wojciechowski, Poland
Henryk Palus, Poland Michał Woźniak, Poland
Wiesław Pamuła, Poland Jan Zabrodzki, Poland
Contents

Computer Graphics, Perception and Image Quality

Hemispherical Gaussians for Accurate Light Integration. . . . . . . . . . . . . . . . 3


Julian Meder and Beat Brüderlin

Gaze-Dependent Screen Space Ambient Occlusion . . . . . . . . . . . . . . . . . . . 16


Radosław Mantiuk

A Fast Algorithm for Quaternion-Based 4D Rotation. . . . . . . . . . . . . . . . . . 28


Aleksandr Cariow, Galina Cariowa, and Dorota Majorkowska-Mech

A Study on Image Comparison Metrics for Atmospheric


Scattering Phenomenon Rendering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
Tomasz Gałaj and Adam Wojciechowski

Graphical Interface Design for Chatbots for the Needs of Artificial


Intelligence Support in Web and Mobile Applications . . . . . . . . . . . . . . . . . 48
Mateusz Modrzejewski and Przemysław Rokita

Algorithms for Random Maps Generation and Their Implementation


as a Python Library . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
Marian Rusek, Rafał Jusiak, and Waldemar Karwowski

Modeling and Rendering of Volumetric Clouds in Real-Time


with Unreal Engine 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
Łukasz Nowak, Artur Bąk, Tomasz Czajkowski,
and Konrad Wojciechowski

Real-Time Simulation of Animated Characters Crowd in Unreal Engine 4 . . . 79


Michał Rosenbeiger, Artur Bąk, and Tomasz Czajkowski

Object Classification and Features

Plane Object-Based High-Level Map Representation for SLAM . . . . . . . . . . 91


Pavel Gritsenko, Igor Gritsenko, Askar Seidakhmet, and Bogdan Kwolek

Level-Set Based Algorithm for Automatic Feature Extraction


on 3D Meshes: Application to Crater Detection on Mars . . . . . . . . . . . . . . . 103
Nicole Christoff, Agata Manolova, Laurent Jorda, Sophie Viseur,
Sylvain Bouley, and Jean-Luc Mari
X Contents

A System for Automatic Town Sign Recognition for Driver


Assistance Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
Dariusz Frejlichowski and Piotr Mikołajczak

Selective and Simple Graph Structures for Better Description of Local


Point-Based Image Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
Grzegorz Kurzejamski and Marcin Iwanowski

Scene Recognition for Indoor Localization of Mobile Robots


Using Deep CNN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
Piotr Wozniak, Hadha Afrisal, Rigel Galindo Esparza,
and Bogdan Kwolek

Character Recognition Based on Skeleton Analysis . . . . . . . . . . . . . . . . . . . 148


Kacper Sarnacki and Khalid Saeed

Weather Characterization from Outdoor Scene Images . . . . . . . . . . . . . . . . . 160


Jenade Moodley and Serestina Viriri

3D and Stereo Image Processing

Clustering Quality Measures for Point Cloud Segmentation Tasks . . . . . . . . . 173


Jakub Walczak and Adam Wojciechowski

Multi-camera Photometric Simulation for Creation of 3D Object


Reconstruction System. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
Dawid Sobel, Karol Jedrasiak, and Aleksander Nawrat

Quality Evaluation of 3D Printed Surfaces Based on HOG Features . . . . . . . 199


Piotr Lech, Jarosław Fastowicz, and Krzysztof Okarma

Convolutional Neural Network-Based Action Recognition on Depth Maps . . . 209


Jacek Trelinski and Bogdan Kwolek

An Integrated Procedure for Calibrating and Distortion Correction


of the Structure Sensor and Stereo-Vision Depth Sensors . . . . . . . . . . . . . . . 222
Dariusz Rzeszotarski and Pawel Strumillo

Second-Order Algebraic Surfaces and Two Image Photometric Stereo . . . . . . 234


Ryszard Kozera and Alexander Prokopenya

Low-Level and Middle-Level Image Processing

Improving RGB Descriptors Using Depth Cues . . . . . . . . . . . . . . . . . . . . . 251


Maciej Stefańczyk
Contents XI

Embedding Spatial Context into Spectral Angle Based Nonlinear


Mapping for Hyperspectral Image Analysis . . . . . . . . . . . . . . . . . . . . . . . . 263
Evgeny Myasnikov

Color Object Retrieval Using Local Features Based


on Opponent-Process Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275
Paula Budzakova, Elena Sikudova, and Zuzana Berger Haladova

Extracting Textual Overlays from Social Media Videos


Using Neural Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287
Adam Słucki, Tomasz Trzciński, Adam Bielski, and Paweł Cyrta

Choosing an Optimal Bracketing Sequence for HDR Imaging. . . . . . . . . . . . 300


Paweł J. Łubniewski and Wojciech S. Mokrzycki

Detection of Pollen Grains in Digital Microscopy Images by Means


of Modified Histogram Thresholding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 308
Dariusz Frejlichowski

Medical Image Analysis

U-CatcHCC: An Accurate HCC Detector in Hepatic DCE-MRI


Sequences Based on an U-Net Framework . . . . . . . . . . . . . . . . . . . . . . . . . 319
Anna Fabijańska, Antoine Vacavant, Marie-Ange Lebre,
Ana L. M. Pavan, Diana R. de Pina, Armand Abergel,
Pascal Chabrot, and Benoît Magnin

Unsupervised Caries Detection in Non-standardized Periapical


Dental X-Rays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 329
Darren Osterloh and Serestina Viriri

Localizing Characteristic Points on a Vertebra Contour


by Using Shape Language . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 341
Marzena Bielecka and Andrzej Bielecki

Lytic Region Recognition in Hip Radiograms by Means


of Statistical Dominance Transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 349
Marcin Kociołek, Adam Piórkowski, Rafał Obuchowicz,
Paweł Kamiński, and Michał Strzelecki

Motion Analysis and Tracking

Aggregation of Binary Feature Descriptors for Compact Scene Model


Representation in Large Scale Structure-from-Motion Applications . . . . . . . . 363
Jacek Komorowski and Tomasz Trzciński
XII Contents

Shallow Convolutional Neural Network and Viterbi Algorithm


for Dim Line Tracking. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 375
Przemyslaw Mazurek

Fast-Tracking Application for Traffic Signs Recognition . . . . . . . . . . . . . . . 385


Abderrahmane Adoui El Ouadrhiri, Jaroslav Burian,
Said Jai Andaloussi, Rachida El Morabet, Ouail Ouchetto,
and Abderrahim Sekkaki

Embedded Vision System for Automated Drone Landing Site Detection . . . . 397
Patryk Fraczek, Andre Mora, and Tomasz Kryjak

Human Face, Gestures and Action Analysis

Silhouette-Based Action Recognition Using Simple Shape Descriptors. . . . . . 413


Katarzyna Gościewska and Dariusz Frejlichowski

Landmark-Based Re-topology of Stereo-Pair Acquired Face Meshes . . . . . . . 425


Eric Patterson, Jessica Baron, and Devin Simpson

A Kinematic Gesture Representation Based on Shape Difference VLAD


for Sign Language Recognition. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 438
Jefferson Rodríguez and Fabio Martínez

Security and Protection

Camera Sensor Traces Analysis in Image Forgery Detection Problem . . . . . . 453


Andrey Kuznetsov

Barcoding in Biometrics and Its Development . . . . . . . . . . . . . . . . . . . . . . 464


Nazym Kaziyeva, Georgy Kukharev, and Yuri Matveev

Digital Image Forensics Technique for Copy-Move Forgery Detection


Using DoG and ORB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 472
Patrick Niyishaka and Chakravarthy Bhagvati

Pattern Recognition and New Concepts in Classification

Does the Research Question Structure Impact the Attention Model?


User Study Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 487
Malwina Dziśko, Anna Lewandowska, and Anna Samborska-Owczarek

Pattern Recognition Method for Classification of Agricultural


Scientific Papers in Polish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 499
Piotr Wrzeciono and Waldemar Karwowski
Contents XIII

Pattern Recognition in Financial Data Using Association Rule . . . . . . . . . . . 512


Krzysztof Karpio and Piotr Łukasiewicz

Ulam Spiral and Prime-Rich Polynomials. . . . . . . . . . . . . . . . . . . . . . . . . . 522


Arkadiusz Orłowski and Leszek J. Chmielewski

Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 535


Computer Graphics, Perception
and Image Quality
Hemispherical Gaussians for Accurate
Light Integration

Julian Meder1,2(B) and Beat Brüderlin1


1
TU Ilmenau, Helmholtzplatz 5, 98693 Ilmenau, Germany
Julian.Meder@tu-ilmenau.de
2
3DInteractive GmbH, Am Vogelherd 10, 98693 Ilmenau, Germany

Abstract. In real-time computer graphics, approximations are often


used to allow for interactive virtual scene rendering. Concerning the
on-line lighting of such scenes, an approach getting increased recogni-
tion is to approximate the light in every direction of the hemisphere of
a surface point using suitable mathematical distribution functions, such
as the well-known Spherical Gaussian. A drawback of this distribution
is that current methods using it are inaccurate and do not reflect the
correct lighting integral over the surface hemisphere. We show new and
more accurate convolution of a Spherical Gaussian with a clamped cosine
distribution. In short, we propose a closed form approximation of the
hemispherical integral of such a distribution in an arbitrary hemisphere.
While our use case is the approximation of the hemispherical lighting
situation, we believe that our general formulation of the hemispherical
integral of a Spherical Gaussian can also be useful in other areas.

1 Introduction

Calculating the lighting of a virtual scene is still one of the most challenging prob-
lems in interactive computer graphics. Here, the idealized expectations of a real-
istic scene clash with the reality of available hardware resources and necessitate
many approximations to at least show plausible results mimicking the physically
accurate. With Spherical Harmonics [9] complex light situations started to get
approximated using mathematical distributions precomputed off-line and evalu-
ated on-line. In recent years, we observe a trend where approximating complex
lighting situations on-the-fly into comparable distributions becomes increasingly
popular. Therefore, we came to study such distributions more closely.
Our main contributions described in this paper include a novel approxima-
tion of the integral of a Spherical Gaussian distribution in a selected hemisphere.
The approximation can be used generally when dealing with these distributions.
Based on this, we additionally propose a more precise convolution of a Spheri-
cal Gaussian with a cosine factor, the latter being often encountered in lighting
calculations. This use case will be explained in more detail in the next section
while the remainder of this paper is organized as follows: Sect. 3 defines Spherical
Gaussians as they are used in this paper and Sect. 4 gives a summary of work
c Springer Nature Switzerland AG 2018
L. J. Chmielewski et al. (Eds.): ICCVG 2018, LNCS 11114, pp. 3–15, 2018.
https://doi.org/10.1007/978-3-030-00692-1_1
4 J. Meder and B. Brüderlin

related to using distributions approximating lighting in general and Spherical


Gaussians in particular. Afterwards, Sect. 5 motivates our new approximations
and derives them. Finally, Sect. 6 shows empirical comparisons and Sect. 7 con-
cludes this paper.

2 Approximating the Rendering Equation

The task of calculating lighting in computer graphics is the solution of the so


called rendering equation [5]:

Lo (x, ωo ) = Le (x, ωo ) + Li (ω)fr (x, ω, ωo )(ω ◦ n)dω (1)
+
Ωn

Essentially, radiance reflected from a surface point x in direction ωo is the self-


emitted radiance Le (x, ωo ), i.e. if x is a light source itself, and the integral over
all possible directions ω of incoming radiance Li (ω) weighted by the bidirectional
reflectance distribution function (BRDF) fr and the cosine between ω and the
surface normal n (◦ being the dot product). We denote the hemisphere over
x in direction n as Ωn+ , i.e. where (ω ◦ n) is greater or equal to 0, Ωn− as the
hemisphere in inverse direction of n, i.e. where (ω ◦ n) is less than 0, and Ω as
the whole sphere.
In our use case we ignore self-emittance Le and approximate (1) as

Lo (x, ωo ) ≈ G(x, ω)(ω ◦ n)dω (2)
+
Ωn

meaning we assume the incoming light times the BRDF can be well approxi-
mated as some distribution G(x, ω), which we chose to be a Spherical Gaussian
and compute on-line from a fully dynamic lighting environment. The remaining
problem is an efficient and accurate convolution of this distribution with the
cosine factor.

3 Spherical Gaussians

A Spherical Gaussian we use in Sect. 2 in the rendering equation is evaluated in


a normalized direction ω and defined as

G(μ, λ, φ, ω) = μeλ(φ◦ω−1) , (3)

where μ is its color value in our case, λ its sharpness and φ its normalized
direction. This distribution has a simple analytical integral over the whole sphere
of possible directions Ω [12]:


μA(λ) = G(μ, λ, φ, ω)dω = μ (1 − e−2λ ). (4)
Ω λ
Hemispherical Gaussians for Accurate Light Integration 5

Multiplication of two Spherical Gaussians G(μ1 , λ1 , φ1 , ω) and G(μ2 , λ2 , φ2 , ω)


simply yields a new Spherical Gaussian G(μ3 , λ3 , φ3 , ω) with

λ3 = λ1 φ1 + λ2 φ2 
λ 1 φ1 + λ 2 φ2
φ3 = (5)
λ3
μ3 = μ1 μ2 eλ3 −λ2 −λ1 .

Respective derivations can be found in [14].

4 Related Work

One of the major foundations concerning the application of mathematical dis-


tributions as the lighting approximation has been established by Sloan et al.
[9] with their work on Spherical Harmonics. The distributions themselves were
precomputed off-line back then, but evaluation during interactive runtime was
very efficient due to the mathematical properties of Spherical Harmonics. They
have since been used many times, recently even completely during runtime in the
works of Kaplanyan et al. [6], using them as representations for light in discrete
volumes of the scene, propagating through the whole volume grid, and Jendersie
et al. [4], approximating the hemispherical light of a discrete set of selected scene
points and interpolating it over the geometry.
In contrast, Spherical Gaussians have been investigated in an interactive
environment by Green et al. [3], who approximated the transport function, i.e.
fr (x, ω, ωo )(ω ◦ n) from Eq. (2), as a weighted mix of Spherical Gaussians. Com-
parable to Sloan et al. [9], Tsai et al. used Spherical Gaussians during a general
evaluation of spherical radial basis functions for precomputed radiance transfer
[13]. Yan et al. found an efficient algorithm to render translucent objects under
Spherical Gaussian Lights accurately [17]. Xu et al. used Spherical Gaussians
to approximate the whole sphere of environmental light during interactive hair
rendering [15]. In similar fashion, Tokuyoshi converted Reflective Shadow Maps
(RSMs) [1] to a small set of Spherical Gaussian Lights and used them to interac-
tively render convincing diffuse and glossy reflections, even supporting caustics
[11]. This work was later extended for more meaningful filtering of the RSM via
kernel-based importance sampling [12].

5 Hemispherical Gaussian Lighting


Because we compute distribution G in (2) dynamically on-line, we decided
against Spherical Harmonics [9] in favor of Spherical Gaussians. Due to the
effective value ranges of Spherical Harmonics being constrained to certain parts
of the unit sphere, a weighted sum of base functions is always used instead of a
single one. While determining the necessary weights is comparatively easy in a
formal sense [8], it involves the evaluation of each chosen base function with each
6 J. Meder and B. Brüderlin

hemispherical light sample. In practice, only a small number of base functions


is used when performing all computations on-line [4,6].
Accumulating the lighting situation into a Spherical Gaussian has been shown
to be more efficient by Tokuyoshi [11] when using Toksvig’s method [10] as
the required costly divisions are independent of the number of hemispherical
samples. Also, with Spherical Harmonics it is difficult to represent distributions
with notably dominant or high frequency directions [3,7]. Spherical Gaussians
in contrast allow to define said dominant direction directly (see Sect. 3).

5.1 Cosine Convolution


Now given the Spherical Gaussian distributions we actually have two problems
while solving Eq. (2):
1. Achieving convolution only around the positive hemisphere Ωn+
2. Accurate integration of a Spherical Gaussian multiplied with a cosine
The first problem is directly given through the rendering Eq. (1). For the second
problem, no analytical solution is known [15]. A common approximation, for
example employed by [12,14,17], is assuming the dot product as a constant and
using the following closed form solution:

G(μ, λ, φ, ω)(ω ◦ n)dω
+
Ωn

(6)
≈ max(0, φ ◦ n) G(μ, λ, φ, ω)dω
Ω
= max(0, φ ◦ n)μA(λ)

Fig. 1. Assuming a constant cosine: (a) Many scattered lights or close area lights create
notable wrong shading compared to (b) the ground truth.

This approximation only works for sufficient sharpness λ, as it implies the direc-
tional distribution of the Spherical Gaussian being narrow enough to be seen as
constant. In other words the distribution is interpreted as a single directional light.
Hemispherical Gaussians for Accurate Light Integration 7

This is not a given in our use case however, as the Spherical Gaussian potentially
represents many lights scattered across the hemisphere or large area light sources
close to the respective pixel’s surface (see Fig. 1).
Another solution proposed by Tokuyoshi [12] is to represent the dot product
as a Spherical Gaussian and integrate over the whole sphere:

G(μ, λ, φ, ω)(ω ◦ n)dω
+
Ωn

≈ G(μ, λ, φ, ω)G(1, 2, n, ω)dω (7)

= G(μ , λ , φ , ω)dω = μ A(λ ).
Ω

1 max(0, cos(β)) 1
cos(β)
λ=1
0.8 SG Cosine
λ = 1.5 0.5

0.6 λ=2
λ=3 0
0.4
−0.5
0.2

0 −1
0 0.7854 1.5708 2.3562 3.1416 0 0.7854 1.5708 2.3562 3.1416
β β
(a) (b)

Fig. 2. Comparing the cosine with its Spherical Gaussian approximations: (a) No mat-
ter the chosen sharpness, a section of the function’s image is either too low or too high
compared to the positive clamped cosine. (b) Our approximation follows the slope of
the non-clamped cosine.

Therefore, the Spherical Gaussian multiplication (5) and subsequent analyt-


ical integral (4) can be used. However, as Fig. 2a shows, the image of the cosine
approximation differs remarkably from the real positive clamped cosine and in
dependence of the chosen sharpness can either be too low or too high, resulting
in overly dark or bright lighting.
We still believe the approximation of the cosine term as a Spherical Gaussian
to be a good start for further investigation. After some experiments, we found
that a Spherical Gaussian with a similar distribution as a cosine is indeed easily
achievable using a modified formulation with an additional parameter α:

G(μcos , λcos , φcos , αcos , ω) = G(μcos , λcos , φcos , ω) − αcos (8)

Automatic non-linear curve fitting this distribution to a standard cosine via Mat-
lab resolved the parameters to μcos = 32.7080, λcos = 0.0315 and αcos = 31.7003
8 J. Meder and B. Brüderlin

with a RMSE of 0.0059. φcos is chosen to be the surface normal n. Figure 2b shows
a comparison of this approximation with the cosine in range [0; π]. With this, we
can now accurately convolve the Spherical Gaussian lighting distribution with a
quasi-cosine analogous to Eq. (7). Integration over Ω would still be easily given by

G(μ, λ, φ, ω)(ω ◦ n)dω
+
Ωn

≈ G(μ, λ, φ, ω)G(μcos , λcos , αcos , n, ω)dω
(9)
Ω 
= G(μ , λ , φ , ω)dω − G(μ, λ, φ, ω)αcos dω
Ω Ω
=μ A(λ ) − μαcos A(λ).

However, as we now convolve with a cosine-like distribution integrating over Ω


creates cancellation effects due to negative cosine values. Our idea here is to
follow the original rendering equation and integrate only in hemisphere Ωn+ . Of
course, this necessitates a solution for integrating a Spherical Gaussian in this
hemisphere.

5.2 Hemispherical Integration


Our findings in the previous section require us to integrate a Spherical Gaussian
in a selected hemisphere. Some specific integrals have simple analytical solutions,
such as its integral in the upper hemisphere Ωφ+
 2π  π
2 2π
Au (λ) = G(λ, φ, ω) sin(θ)dθdϕ = (1 − e−λ ), (10)
0 0 λ

in the lower hemisphere Ωφ− rotated 180◦ away from φ


 2π  π
2π −2λ λ
Ab (λ) = G(λ, φ, ω) sin(θ)dθdϕ = e (e − 1) (11)
0 π
2
λ

and in the hemisphere rotated by 90◦ away from φ


 π π
π
Am (λ) = G(λ, φ, ω) sin(θ)dθdϕ = (1 − e−2λ ) (12)
0 0 λ
We obtained these solutions by using symbolic integration via Matlab. μ will
be omitted for all further derivations as by its constant nature it can be pulled
out of the integral immediately. For these and all following equations we denote
 T
ω = sin(θ) cos(ϕ), sin(θ) sin(ϕ), cos(θ) in spherical coordinates.
Generally, we can express the integral over some hemisphere rotated away
from the original upper hemisphere by rotating ω in (10) by some angle β.
In our use case, β is given by acos(φ ◦ n), i.e. the angle between the Spherical
Gaussian’s direction and the surface normal. As Spherical Gaussians are isotropic
Hemispherical Gaussians for Accurate Light Integration 9

6
1
λ = 0.1

h
5

Normalized integral A
0.8 λ = 10
λ = 500
Integral Ah

4 λ = 0.1
λ = 10 0.6
3
λ = 500
0.4
2

1 0.2

0 0
0 0.7854 1.5708 2.3562 3.1416 0 0.7854 1.5708 2.3562 3.1416
Hemisphere angle β Hemisphere angle β
(a) (b)

Fig. 3. Plots of numerical integration values Ah in dependence on β: (a) Absolute


values and (b) curves normalized to [0; 1].

in distribution, we assume w.l.o.g. that the rotation occurs around the local x
axis. Therefore the rotation is given through the matrix
⎛ ⎞
1 0 0
M (β) = ⎝0 cos(β) sin(β) ⎠ (13)
0 − sin(β) cos(β)

Inserting this rotation into (10) and w.l.o.g. assuming φ = (0, 0, 1)T yields
 2π  π
2
Ah (λ, β) = G(λ, (0, 0, 1)T , M ω) sin(θ)dθdϕ
0 0
 2π  π
2
(14)
λ(cos(β) cos(θ)−sin(β) sin(φ) sin(θ)−1)
= e sin(θ)dθdϕ
0 0

Unfortunately, we did not find a direct analytical solution for Ah . We there-


fore investigated the integral numerically to look for possible closed form approx-
imations. When plotting the discrete values of Ah in relation to β as in Fig. 3a
and additionally normalizing the curves to an image range of [0; 1] as in Fig. 3b
some interesting observations can be made:

– The integral values vary smoothly and decrease monotonically from Au (β =


0) to Ab (β = π) passing through Am when β = π2 .
– The function shape appears to be dependent on λ. It becomes cosine-like as
λ approaches zero and step-like, when λ approaches infinity.
– The function seems to be mirrored by an axis passing through Am .

These observations substantiate the suspicion of a logistic curve [2] for us. We
therefore hypothesize, that the integral of a Spherical Gaussian over an arbi-
trary hemisphere can be well approximated by calculating Au and Ab following
10 J. Meder and B. Brüderlin

Eqs. (10) and (11) and interpolating between these values using

Ah (λ, β) ≈ Au (λ)s(λ, β) + Ab (λ)(1 − s(λ, β)), (15)

with s being a logistic curve over cos(β) with steepness t(λ). As such a curve does
not have an image of [0; 1] in input range [0; π] in general, which is necessary for
interpolation (15), we additionally normalize it:
1
s (λ, β) =
ecos(β)t(λ)
+1
(16)
s (λ, β) − s (λ, π)
s(λ, β) = 
s (λ, 0) − s (λ, π)

For ease of computation on the GPU (16) can be rewritten as

et(λ) et(λ) cos(β) − 1


s(λ, β) =
(et(λ) − 1)(et(λ) cos(β) + 1)
(17)
ab − 1
= , a = et(λ) , b = et(λ) cos(β) .
(a − 1)(b + 1)

50

40
Steepness t(λ)

30

20 Fitted values
Our approximation
10

0
0 100 200 300 400 500 600 700
Sharpness λ

Fig. 4. Curve fitted steepness values t(λ) compared to our closed form approximation

With this approximation, we were able to accurately fit the hemispherical


integrals depending on λ and β with RMSEs given in Sect. 6.1.
What remains is a closed form approximation for the steepness t(λ). Again,
looking at the values resulting from non-linear curve fitting shown in Fig. 4 we
recognize the familiar shape of a square root. Indeed, different modifications of
the square root gave us good fits to the curve in Fig. 4 and we ultimately used
the solution given in the next section.
Hemispherical Gaussians for Accurate Light Integration 11

Algorithm 1. Approximate Hemispherical Gaussian Convolution


function Ah (λ, cos(β))
√ 2
t ← λ λ1.6988λ +10.8438λ
2 +6.2201λ+10.2415
t
a←e
b ← et cos(β)
ab−1
s ← (a−1)(b+1)
return Ab (λ)(1 − s) + Au (λ)s
end function

function SG Cosine Convolve(μ, λ, φ, n)


μcos ← 32.7080
λcos ← 0.0315
αcos ← 31.7003
λ ← λcos n + λφ
φ ← λcosλn+λφ


μ ← μcos μeλ −λ−λcos
return μ Ah (λ , φ ◦ n) − μαcos Ah (λ, φ ◦ n)


end function

5.3 Final Algorithms

Using the findings of the previous section, the cosine approximation (8) and the
Spherical Gaussian multiplication (5) we can now closely approximate convolu-
tion (2). Algorithm 1 summarizes both this and our method for calculating the
general hemispherical integral Ah .
Clearly, our approach is more costly than the approximations we referred to in
Sect. 5.1, involving a total of 4 individual Spherical Gaussian integral evaluations
plus the additional overhead for the two hemispherical approximations. However,
as the next section will prove empirically, it carries more overall accuracy and
shows a more generalized behavior when applying it to our use case. Also, as
Au (λ) and Ab (λ) are mainly composed of identical factors Algorithm 1 can be
heavily optimized to avoid unnecessary recalculations.

6 Evaluation

6.1 Approximation Error


In this section, we evaluate our and other approximations against the ground
truth given by numerical integration over β ∈ [0; π] for a range of different
sharpness values λ. We both show the absolute error between the curves and a
relative error by normalizing the curves to an image range of [0; 1], the latter
giving a better comparison of the integral shapes.
Figure 5 shows the positive result of using our accurate cosine representation
combined with our novel hemispherical integration. While the other approxima-
tions show significantly increasing error values, especially when λ approaches 0,
12 J. Meder and B. Brüderlin

0.2
Our, absolute
0.175 Our, normalized
0.15 Approx. Cosine, absolute
0.125 Approx. Cosine, normalized

RMSE
Const. Cosine, absolute
0.1 Const. Cosine, normalized
0.075
0.05
0.025
0 0 1 2
10 10 10
Sharpness λ

Fig. 5. Comparison between the RMSE values to the ground truth numerical inte-
gral when using either of the two existing cosine convolution approximations or our
approximation (see Sect. 5.1).

0.01

0.008
Absolute
RMSE

0.006 Normalized

0.004

0.002

0 0 1 2
10 10 10
Sharpness λ

Fig. 6. RMSE values for our hemispherical integral approximation compared to the
numerical integral.

our approximation remains a good fit overall. Our method additionally matches


the shape of the integral closely for the most part, as is evident when comparing
the normalized curves.
We did not find any other works proposing a general approximation for the
integral of a Spherical Gaussian over an arbitrary hemisphere. Therefore, for the
sake of completeness we include the RMSEs of our general hemispherical integral
approximation in Fig. 6 for further reference.

Fig. 7. (a) Under conditions of broad hemispherical lighting, the assumption of the
constant cosine breaks down, but (b) the errors of the approximate cosine are not
easily noticeable in this case. (c) Our solution is a close match to (d) the brute force
ground truth. (Bunny model Stanford
c University)
Hemispherical Gaussians for Accurate Light Integration 13

Fig. 8. (a) Assuming the cosine factor to be constant works in situations of narrow
hemispherical lighting while (b) using the approximate cosine wrongly shows brighter
results. (c) Again, our solution is a close match to (d) the brute force ground truth
(Bunny model Stanford
c University).

6.2 Visual and Speed Comparison


For the visual comparison, we use gathering on a Reflective Shadow Map [1] into
a Spherical Gaussian distribution per surface point. The final lighting is then
calculated using either of the approximations from Sect. 5.1.
The approximation of the rendering integral as in Eq. (6), assuming the con-
stant cosine, behaves as expected from its underlying assumptions: It shows good
results when the lighting situation is narrowly distributed over the hemisphere
(Fig. 8a) and unsuitable results for broad hemispherical lighting (Fig. 7a). The
less accurate Spherical Gaussian cosine approximation of Tokuyoshi [12] shows
brighter results than the ground truth in both situations, while perceptively
being a little less recognizable in the broad lighting situation (see Figs. 7b and
8b), as the longer tail of the distribution causes higher integral values.
Comparing these with our approximation in Figs. 7c and 8c shows results
close to the ground truth solution in Figs. 8d and 7d. This yields the conclusion
that we propose a more generalized solution for this particular use case.

Table 1. Computation time per frame averages (ms)


Method Resolution
1280 × 720 1920 × 1080 2560 × 1600 3840 × 2400
Constant Cosine 4.90 5.56 6.30 10.59
Approximate Cosine 5.07 5.78 6.73 11.70
Ours 5.29 6.22 7.43 13.83

Our qualitative comparison in Sect. 5.3 already predicted that our increased
precision comes at the cost of performance. However, as Table 1 shows our testing
environment remains real-time capable when we replace the previous methods
with our novel approach. Indeed, the increase in processing time is negligible
compared to the removal of the noticeable visual errors of the other methods.
14 J. Meder and B. Brüderlin

7 Conclusion
Over the course of this paper we developed both a more accurate way to convolve
a Spherical Gaussian with a cosine factor and a novel general approximation to
the problem of integrating a Spherical Gaussian over an arbitrary hemisphere.
While our solution is little more computationally expensive, its properties allow
a more generalized application. In addition, our generalized hemispherical inte-
gration method is not constricted to problems in computer graphics, but may
be applied in other areas where Spherical Gaussians are employed.
For future work, we will investigate a direct closed form approximation of
integral (2), as this can yield a cheaper and potentially even more accurate solu-
tion. Furthermore, we find the idea of integrating anisotropic Spherical Gaussians
[16] over arbitrary hemispheres to be a very interesting research direction.

References
1. Dachsbacher, C., Stamminger, M.: Reflective shadow maps. In: Proceedings of the
2005 Symposium on Interactive 3D Graphics and Games, pp. 203–231. I3D 2005.
ACM, New York (2005). http://doi.acm.org/10.1145/1053427.1053460
2. Garnier, J., Quetelet, A.: Correspondance mathématique et physique. No.
10, Impr. d’H. Vandekerckhove (1838). https://books.google.de/books?
id=8GsEAAAAYAAJ
3. Green, P., Kautz, J., Matusik, W., Durand, F.: View-dependent precomputed light
transport using nonlinear Gaussian function approximations. In: Proceedings of
the 2006 Symposium on Interactive 3D Graphics and Games, pp. 7–14, I3D 2006.
ACM, New York (2006). http://doi.acm.org/10.1145/1111411.1111413
4. Jendersie, J., Kuri, D., Grosch, T.: Precomputed illuminance composition for real-
time global illumination. In: Proceedings of the 20th ACM SIGGRAPH Symposium
on Interactive 3D Graphics and Games, pp. 129–137, I3D 2016. ACM, New York
(2016). http://doi.acm.org/10.1145/2856400.2856407
5. Kajiya, J.T.: The rendering equation. SIGGRAPH Comput. Graph. 20(4), 143–
150 (1986). http://doi.acm.org/10.1145/15886.15902
6. Kaplanyan, A., Dachsbacher, C.: Cascaded light propagation volumes for real-time
indirect illumination. In: Proceedings of the 2010 ACM SIGGRAPH Symposium
on Interactive 3D Graphics and Games, pp. 99–107, I3D 2010. ACM, New York
(2010). http://doi.acm.org/10.1145/1730804.1730821
7. Ng, R., Ramamoorthi, R., Hanrahan, P.: All-frequency shadows using non-linear
wavelet lighting approximation. ACM Trans. Graph. 22(3), 376–381 (2003).
http://doi.acm.org/10.1145/882262.882280
8. Sloan, P.P.: Stupid spherical harmonics (SH) tricks. In: GDC 2008 (2008)
9. Sloan, P.P., Kautz, J., Snyder, J.: Precomputed radiance transfer for real-time
rendering in dynamic, low-frequency lighting environments. ACM Trans. Graph.
21(3), 527–536 (2002). http://doi.acm.org/10.1145/566654.566612
10. Toksvig, M.: Mipmapping normal maps. J. Graph. Tools 10(3), 65–71 (2005).
https://doi.org/10.1080/2151237X.2005.10129203
11. Tokuyoshi, Y.: Virtual spherical Gaussian lights for real-time glossy indirect illu-
mination. Comput. Graph. Forum 34(7), 89–98 (2015). http://dx.doi.org/10.1111/
cgf.12748
Hemispherical Gaussians for Accurate Light Integration 15

12. Tokuyoshi, Y.: Modified filtered importance sampling for virtual spherical Gaussian
lights. Comput. Vis. Media 2(4), 343–355 (2016). https://doi.org/10.1007/s41095-
016-0063-3
13. Tsai, Y.T., Shih, Z.C.: All-frequency precomputed radiance transfer using spherical
radial basis functions and clustered tensor approximation. In: ACM SIGGRAPH
2006 Papers, pp. 967–976, SIGGRAPH 2006. ACM, New York (2006). http://doi.
acm.org/10.1145/1179352.1141981
14. Xu, K., Cao, Y.P., Ma, L.Q., Dong, Z., Wang, R., Hu, S.M.: A practical algo-
rithm for rendering interreflections with all-frequency BRDFS. ACM Trans. Graph.
33(1), 10:1–10:16 (2014). http://doi.acm.org/10.1145/2533687
15. Xu, K., Ma, L.Q., Ren, B., Wang, R., Hu, S.M.: Interactive hair rendering and
appearance editing under environment lighting. ACM Trans. Graph. 30(6), 173:1–
173:10 (2011). http://doi.acm.org/10.1145/2070781.2024207
16. Xu, K., Sun, W.L., Dong, Z., Zhao, D.Y., Wu, R.D., Hu, S.M.: Anisotropic spherical
Gaussians. ACM Trans. Graph. 32(6), 209:1–209:11 (2013). http://doi.acm.org/
10.1145/2508363.2508386
17. Yan, L.Q., Zhou, Y., Xu, K., Wang, R.: Accurate translucent material render-
ing under spherical Gaussian lights. Comput. Graph. Forum 31(7pt2), 2267–2276
(2012). http://dx.doi.org/10.1111/j.1467-8659.2012.03220.x
Gaze-Dependent Screen Space Ambient
Occlusion

Radoslaw Mantiuk(B)

West Pomeranian University of Technology,


Szczecin al. Piastów 17, 70–310 Szczecin, Poland
rmantiuk@zut.edu.pl

Abstract. The screen space ambient occlusion (SSAO) is a fast global


illumination technique, which approximates interreflections between ren-
dered objects. Due to its simplicity, it is often implemented in commercial
computer games. However, despite the fact that SSAO calculations take
a few milliseconds per frame, a significant computation load is added
to the total rendering time. In this work we propose a technique, which
accelerates the SSAO calculations using information about observer’s
gaze direction captured by the eye tracker. The screen region surround-
ing the observer’s gaze position is rendered with maximum quality, which
is reduced gradually for higher eccentricities. The SSAO quality is vary-
ing by changing the number of samples that are used to approximate
the SSAO occlusion shadows. The reduced sampling results in almost
two-fold acceleration of SSAO with negligible deterioration of the image
quality.

1 Introduction
In the Phong reflection model, diffuse and specular reflections are varying due
to observer and lights positions, but ambient light is constant. Having this
assumptions, we miss the interreflections between rendered objects. Adding
ambient occlusion (AO) for varying ambient light creates very convincing soft
shadows, that combined with direct lighting give realistic images [1, Sect. 9.2].
The AO technique is faster in comparison to the full global illumination solu-
tions, however, it still needs demanding resources to achieve high quality render-
ings. An approximation of this technique, called screen space ambient occlusion
(SSAO) [15] simulates local occlusions in real time. However, the accuracy of
approximation strongly depends on the sampling density, which in turn limits
its applications. It is especially a drawback in the game engine, in which the ren-
dering time spent for AO computation should use only a fraction of the frame
time because other calculations determine the quality of the gameplay. Thus,
further acceleration of the SSAO computations is a desirable task which has
significant impact on the overall quality of the real-time computer graphics.
In this work we present a gaze-dependent screen space ambient occlusion
technique in which information about human viewing direction is employed
to vary accuracy of the occlusion factors. The screen region surrounding the
c Springer Nature Switzerland AG 2018
L. J. Chmielewski et al. (Eds.): ICCVG 2018, LNCS 11114, pp. 16–27, 2018.
https://doi.org/10.1007/978-3-030-00692-1_2
Gaze-dependent Screen Space Ambient Occlusion 17

observer’s gaze position is rendered with maximum precision, decreasing gradu-


ally towards parafoveal and peripheral regions. The idea of this solution is based
on the directional characteristic of the human visual system (HVS). People see
the high frequency details only in a small viewing angle subtended 2–3◦ of the
field of view. In this range, people see with a resolution of up to 60 cycles per
angular degree, but for a 20-degree viewing angle, this sensitivity is reduced even
ten times [8].
The accuracy of the occlusion factors is determined by the number of samples
used to calculate these factors. In practice, less samples results in ragged edges of
the SSAO shadows. However, this aliasing artifacts are barely visible by humans
in peripheral regions of vision. In other words, image deterioration caused by low
sampling in the SSAO technique can be clearly visible only in high frequency
regions. Number of samples in SSAO can be gradually reduced with distance
to a gaze point what significantly speeds-up rendering. We use the eye tracker
to capture the human gaze point. Then, sampling is reduced with eccentricity
(deviations from the axis of vision) along the curve determined by the gaze-
dependent contrast sensitivity function (GD-CSF) [5]. This perceptual function
models loss of contrast sensitivity with eccentricity. It can be used to determine
the maximum special frequency visible for humans for an arbitrary viewing angle.
Section 2 gives background information on the screen-space ambient occlu-
sion technique and outlines the previous work. Section 3 is focused on our gaze-
dependent extension of SSAO and shows how sampling can be reduced without
noticeable image deterioration. Section 4 presents the results of the perceptual
experiment in which we evaluate the perceptual visibility of the image deterio-
rations.

2 Background and Previous Work

Screen space ambient occlusion (SSAO) is a rendering technique for approxi-


mating the global illumination in real time [15]. For every pixel on the screen,
the depths of surrounding pixels are analyzed to compute the amount of occlu-
sion, which is proportional to the depth difference between a current pixel and
a sampled pixel.
We implemented the normal-oriented hemisphere SSAO technique [20]. The
hemisphere is oriented along the surface normal at the pixel. The samples from
the hemisphere are projected into screen space to get the coordinates into the
depth buffer. If the sample position is behind this sample depth (i.e. inside
geometry), it contributes to the occlusion factor. The procedure is repeated
for every pixel in image to generate the map of occlusion factors (also called
the occlusion shadows, see example in Fig. 1). This map forms a characteristic
shadowing of ambient light, which is visible as a high frequency information
in characteristic regions of the scene (e.g. at corners, close to complex objects,
etc.). The shadows are blended with the pixel colors computed based on the
Phong lighting equation. Wherein, frequency of the shadows is often higher than
variability of the Phong ambient shading.
Another random document with
no related content on Scribd:
voorbehouden hebben zelf te oordeelen over den spoed van de
verschillende werkzaamheden, doch tegenwoordig luisterde dat zoo
nauw niet.

„Roep me dan tegen vier uren,” zei hij.

„Als je dan maar dadelijk opstaat, anders ga ik alleen.”

Hij lachte er om.

„Nu, ik zie jou al wandelen, alleen!”

Lucie was zenuwachtig; toen hij zijn kamer was binnengegaan, liep
ze doelloos rond in de achtergalerij; nu en dan rolden haar tranen
langs de wangen, die ze snel afwischte, soms met haar zakdoek,
dan weer gedachteloos met een punt harer kabaja; tegen halfvier
baadde zij, en daarna wekte ze haar man. Van Brakel stond dadelijk
op, greep, zooals hij dat gewoon was, een handdoek en stormde
met verwilderde haren en een rood gezicht naar de badkamer, waar
hij dan wel een half uur bleef rondplassen om zich te verfrisschen en
de laatste v a p e u r s van de ochtend-bittertjes te doen verdwijnen.

Toen hij terugkwam, wandelde Lucie reeds gekleed op het voorerf.


[114]

„Wat heb je toch ’n haast,” riep hij haar toe. „’t Is pas kwart voor
vijven.”

„Ja, in den donker wandel ik niet.”

„Nu, ik zal me ook haasten, hoor! In een wip ben ik gekleed.”

Dat was hij ook, maar ten koste van menigen zweetdroppel, en toen
hij naar buiten kwam, scheen het effect van zijn bad reeds geheel
verloren.
„Het is gruwelijk warm de laatste dagen.”

„Ja, het is erg warm.”

„Het baden brengt ook geen verfrissching.”

„Neen, het water is niet eens koel.”

’t Was het oude, eeuwige thema van Europeanen in Indië: de


warmte, en nog eens de warmte. Zoo spraken zij al voortwandelend
over koetjes en kalfjes, en hij lette er niet op, dat zij intusschen de
richting aangaf, en nu eens rechts, dan links omsloeg, terwijl hij
achteloos meeliep.

Plotseling stond hij stil. Als ze doorliepen, kwamen zij voorbij het
huis van mevrouw Du Roy, en dat wilde hij niet. Het zou te gek zijn!
Hij wist bovendien, dat als bij toeval die „dame” aan het hek stond,
zij in staat was de een of andere familiariteit te begaan.

„We moesten maar teruggaan.”

„Waarom?”

„Wel, we zijn al een heel eind opgeloopen; het is zoo warm.”

„We kunnen nog best een eindje wandelen.”

„Pff!… God, het is zoo vreeselijk benauwd, Luus! Laat ons toch
liever teruggaan.”

„Kom, kom! zoo erg is het niet. Het schijnt wel, dat [115]je altijd alleen
naar huis verlangt als je met mij uitgaat.”

Na frappeerde hem wel die zonderlinge toon, waarop zij sprak en die
haar anders nooit eigen was, maar aan de ware reden dacht hij zelfs
nu nog niet. Hij had zich een min of meer bepaalde voorstelling
gemaakt van een mogelijke ontdekking door Lucie van zijner relatie
met mevrouw Du Roy; daar kwam een s c è n e van, meende hij, met
tranen en verwijten, met de bedreiging, dat zij van hem wilde
scheiden, en verder met een eindeloos twisten en standjes maken.

Hij kon nu moeilijk weigeren verder te gaan, indien zij er zoo op


gesteld was.

„Hoe flauw!” zei hij knorrig. „Alsof ik niet altijd bereid ben met je uit te
gaan, wanneer en zoolang je maar wilt.”

Zij klemde de lippen opeen, zeide niets en ging verder. Van Brakel
keek scherp door zijn bril. Goddank, mevrouw Du Roy stond niet aan
den ingang van haar erf; als ze er nu maar voorbij waren, was het
niets; dan zou hij met Lucie den anderen kant om naar huis gaan.

Plotseling ontstelde hij. Ze waren nog slechts een paar huizen van
het nette erf der huizinge Du Roy verwijderd, toen plotseling zijn
hond, die hem als gewoonlijk ook nu was gevolgd, vooruitschoot,
voor den ingang van het erf stilstond, en kwispelstaartend zijn
meester aanzag, met een uitdrukking in den kop en in de geheele
houding, welke volkomen en onmiskenbaar het karakter droeg eener
uitnoodiging om binnen te gaan.

„Hier!” bulderde Van Brakel, harder dan noodig en netjes is, als men
met een dame wandelt. „Hier!”

Doch de hond stoorde zich er niet aan. ’t Huis, waar hij [116]voor
stond, scheen hem lief, en hij holde ondanks de woede zijns
meesters het erf over en de trappen van de voorgalerij op.

Lucie was krijtwit geworden. Zij keek het erf op toen ze er


voorbijgingen; ze zag in de voorgalerij een vrouw in sarong en
kabaja, die op haar hurken zat en den hond streelde, en ze zag, dat
die vrouw den eenen arm in de hoogte stak en met de vingers
groette op Indische manier.

Er werd verder geen woord tusschen Van Brakel en zijn vrouw


gewisseld. Hij scheen nu te begrijpen, dat zij alles reeds wist en dat
ze niet bij toeval dien weg had gekozen. Nu en dan bekeek hij haar
van ter zijde, en dan deed hem die smartelijke trek op haar gezicht
pijnlijk aan. Hij durfde niets te zeggen. Zoo kwamen ze thuis, en toen
zij in de kamer ging om van kleeren te verwisselen, bleef hij achter,
besluiteloos, niet wetende wat aan te vangen. Hij hoorde zijn vrouw
de bedienden roepen, en bespeurde een op dat uur van den dag
bijzonder groote bedrijvigheid, welke niets goeds voorspelde. Zou ze
heengaan en hem in den steek laten? Zou het een publiek
schandaal worden? Na eenige angstige oogenblikken zag hij, dat het
zoo’n vaart niet loopen zou: de logeerkamer werd in orde gebracht.
Wie de l o g é zou zijn, begreep hij. Zijn positie in de huishouding
werd daardoor voorloopig als vanzelf aangewezen. Hij zag zijn
kleeren de eene kamer uit- en de andere indragen. ’t Hinderde hem
zoo als ’t ware de deur te worden uitgezet in zijn eigen huis, maar hij
dacht niet aan eenig verzet. Als zij het dan toch te weten moest
komen, was het maar beter zóó, vond hij. ’t Was wel hoogst
onaangenaam, maar in ’s Hemels naam, er was niets aan te doen.
Hij kon immers [117]niet anders! Indien hij die vervloekte schuld niet
aan mevrouw Du Roy had gehad, dan zou hij.… och, maar dan zou
er immers nooit iets zijn gebeurd, want eigenlijk hield hij alleen van
Lucie, en kon die andere hem hoegenaamd niets schelen. Het was
niet anders. Hij moest varen in het schuitje, waarin hij zat; ’t was
o n m o g e l i j k iets daartegen te doen, hoe miserabel het ook mocht
zijn; à l ’ i m p o s s i b l e , dacht hij, n u l n ’ e s t t e n u .

En daar ophelderingen toch tot niets konden leiden, en hij geen


beterschap kon beloven, ging hij, ten einde buiten alle discussie te
blijven, maar weer de deur uit, tot het tijd was om te eten.
Geerling had het wel bemerkt. Hij zag zeer goed dat Van Brakel zich
onmenschelijk inspannen moest om aan tafel een vroolijken, lossen
toon aan te slaan; dat mevrouw Van Brakel koel en kort tegen haar
man sprak, maar tegen hem, Geerling, allervriendelijkst en
voorkomend was; dat Van Brakel dit laatste met geheimen angst en
tegenzin gadesloeg.

Het gebeurde dikwijls tegenwoordig, dat Geerling samen met Van


Brakel uitging, maar het jonge mensch ging niet meer naar de
woning van mevrouw Du Roy; hij ging naar Ceciel, en zat daar tot elf
uren, met het meisje in de binnengalerij, in tegenwoordigheid harer
moeder, die een roman las.

In den laatsten tijd had hij zich zelfs meermalen ongunstig over
mevrouw Du Roy uitgelaten, en half en half te kennen gegeven, dat
hij Van Brakel’s handelwijze afkeurde.

„Is ze er achter?” vroeg hij, toen ze buiten kwamen.

„Het schijnt zoo.” [118]

„Heeft ze dan niets gezegd?”

„Neen, maar als ik je alles vertel, zal je wel kunnen begrijpen, dat ze
’t weten m o e t .”

„Drommels, ja,” antwoordde Geerling ernstig. „Dat ziet er gek uit.


Een vrouw, die het zóó opneemt, is, dunkt me, veel erger, dan een,
die s c è n e s maakt.”

Toen Van Brakel zweeg, vervolgde hij:

„Laat ik je een raad geven.”


„Och, wat.…”

„Wezenlijk; je zult me toch niet wijsmaken, dat je die moeder Du Roy


prefereert boven je eigen vrouw?”

„Neen, waarachtig niet.”

„Welnu, wees dan ook niet dom. Ik vind er niets in, dat een getrouwd
man eens een dwaasheid doet. Men kan niet altijd braaf wezen!
Maar laat het gekheid blijven, en niet tragisch worden. Daarvoor
moet je zorgen.”

„Ik kan niet.”

„Nonsens!”

„Dat is geen n o n s e n s ! Enfin, ik kan het je niet zeggen, maar laat


me asjeblieft met rust. Het is waarachtig al beroerd genoeg.”

Geerling haalde de schouders op. Bij zulk een humeur was het maar
beter te zwijgen, en met minachting voor den man, die zoo miserabel
onder de plak zat van twee vrouwen, vertelde hij Ceciel het gesprek.

„Ik wist het,” zei ze. „Hij heeft gelijk; hij kan niet. Zij heeft finantiëel
hem geheel in haar macht.”

„Zoo, is dat de quaestie?”—Hij dacht even na. Zou het haar


misschien genoegen doen, als hij zich royaal toonde; [119]zou ze
daardoor meer van hem houden? Want daaraan was hem veel
gelegen, en in zoover was hij gelijk aan Van Brakel, dat ook hij onder
de plak zat, wel op een geheel andere wijze, maar niettemin er
onder. Hij had reeds de toestemming harer ouders, en hij had
geschreven naar Holland, zoo gedecideerd, meende hij, dat zijn
ouders wel zouden inzien, hoe weinig zij vermochten, indien zij
wellicht bezwaren of andere plannen hadden.
„Zou het veel zijn?” vroeg hij voorzichtigheidshalve.

„Ik weet het niet; ik spreek tante in den laatsten tijd haast niet; naar
vroegere uitlatingen van haar te oordeelen, is het wel een duizend of
tien.”

„Als ik Van Brakel eens trachtte te helpen?”

Ceciel dacht een oogenblik na. Als haar aanstaande man—want ze


was zeker, dat ze hem daarvoor kon houden—Van Brakel hielp,
vloeide het geld terug in de kas van tante, wier erfgename zij
vermoedelijk was. Daartegenover stond, dat Geerling het geld kwijt
was, want dat een speler als Van Brakel het ooit zou terugbetalen
geloofde zij niet.

„Wel,” zei ze, „ik raad het je niet aan. De man is niet te helpen, en
tante Du Roy is wel waard, dat ze eens een grooten klap krijgt.
Misschien leert het haar die schandelijke caprices af.”

„Volkomen juist. Ik zal het dus maar op zijn beloop laten.”

Eigenlijk was hij er blij om. Wat was zij toch een heerlijk meisje. Men
kon h a a r gerust voorstellen iets dwaas te doen; zij was veel te
verstandig het te permitteeren.

Hij had elke gedachte aan iets anders dan een geregelde,
fatsoenlijke verstandhouding tusschen hem en haar laten [120]varen.
Ceciel had hem klaar en duidelijk aan het verstand gebracht, dat zoo
iets met h a a r onmogelijk was, en Geerling was zóó verliefd geraakt
dat hijzelf in staat ware geweest die onmogelijkheid onder eede te
bevestigen en het in vrede of in strijd op te nemen tegen iedereen,
die het niet met hem eens was. In deze Don-Quichotte-periode ging
het hem, al was hij in Holland nog zoo „uit”-geweest, precies als
ieder ander.
„Ik wou dat het antwoord maar kwam,” zei hij ongeduldig.

„Ik niet,” antwoordde zij met een zucht, die hem streelde.

„Hou je zoo weinig van me?”

„Dat weet je wel beter, maar ik vrees dat je oudelui hun toestemming
niet zullen geven.”

„Och kom! Wat kunnen ze er tegen hebben?”

„Heel veel. Tegen mij persoonlijk natuurlijk niets; ze kennen me niet


eens; tegen papa en mama ook niet, tenzij je ouders een r i j k e
familie verlangen.”

„Welnu?”

„Het is alweer tante Du Roy. Je oudelui en andere leden in je familie


hebben hier op de plaats gewoond, en natuurlijk kennen ze haar van
haar ongunstige zijde.…”

„Maar lieve, beste Ceciel, wat hebben ze toch met dat mensch
noodig! Ik heb haar naam natuurlijk niet genoemd. Waarschijnlijk
weten ze niet eens, dat je aan die vrouw bent geparenteerd.”

„Alsof hun dit niet van hier zal geschreven worden! Denk je
misschien, dat men op je kantoor niet weet, hoe dikwijls je ’s avonds
hier komt? Hoeveel mailbrieven zouden er al verzonden zijn naar
Holland met particuliertjes over ons?” [121]
Geerling kon het niet ontkennen. ’t Was volkomen waar, dacht hij,
maar ’t was sterk, dat ze gedacht had aan zijn kantoor en aan de
mailbrieven.

„Nu,” antwoordde hij, „het moge zijn, zoo het wil. Wij trouwen met de
toestemming en anders zonder.”

„Het zou me zoo’n leed doen,” zei ze in vollen ernst, de handen over
elkaar leggend in de schoot, en met haar mooie oogen
omhoogblikkend naar ’t plafond, „het zou me zoo innig spijten,
wanneer we op die manier moesten trouwen. Ik geloof wel, dat het
zou voorbijgaan, maar toch had ik het liever anders.”

„Maar lieve Ceciel, er is immers nog niets.…”

„Het z a l komen, dat voel ik. Ik ben zeker, dat het komen zal. Wie
weet hoe? Wie weet hoe ik belasterd ben en wat je over me zult te
hooren krijgen.”

Hij nam een harer handen in de zijne en zwoer, dat wat het ook zijn
mocht en hoeveel kwaad ook van haar werd gesproken, het nimmer
in hem zou opkomen er een letter van te gelooven. Hij kende haar
immers goed!

„Dat is het niet, Jules,” ging zij voort op denzelfden toon. „Als ik
dacht dat jij ’t zoudt gelooven, zou ik je niet willen hebben; als ik niet
overtuigd was van jou goede trouw zou ik je de deur wijzen. Eens en
vooral laat dat denkbeeld zelfs nimmer tusschen ons beproken
worden. Ik heb je alles gezegd, en ik verlang je vertrouwen, omdat ik
er recht op heb; het idee alleen van wantrouwen is beleedigend.”

„Lieve Ceciel, er is immers geen haar op mijn hoofd, dat er aan


denkt. Ik zei het alleen.…”
„O ja, ik weet het immers wel! Maar wat ik van jou verg, [122]kan ik
niet eischen van je familie; daar trouw ik wel in, maar ik t r o u w haar
niet.”

„Laat ons geen zorgen hebben vóór den tijd. Wij weten er immers
nog niets van.”

Zij zweeg en glimlachte; zij wist het immers heel goed! Het antwoord
kon niet goed uitvallen; zij zou haar positie in de familie Geerling
moeten overwinnen; en dat zou ze, dáár voelde zij zich volmaakt
voor berekend. Het eenige, wat haar metterdaad hinderde, was het
uitstel, dat een weigering onvermijdelijk zou teweegbrengen. Uitstel
leidde zoo gemakkelijk tot afstel, en Geerling, hoe verliefd ook, was
geen man van geestkracht en volharding; dat zag ze heel goed in.

Dien avond reed juist voor Geerling’s voertuig een andere wagen het
erf op bij Van Brakel, wat den commensaal zeer interesseerde; met
nog meer belangstelling zag hij er de vrouw des huizes uitstappen,
heel alleen.

„U mevrouw?” vroeg hij. „Nog zoo laat uit geweest?”

„Ja,” antwoordde zij kortweg. „Ik wenschte u tevens nu maar te


zeggen, wat ik anders tot morgenochtend had aangehouden. U zult
wel zoo goed willen zijn zoo spoedig mogelijk een andere woning te
zoeken.”

Hij stond er verstomd van.

„U is wel vriendelijk,” antwoordde hij geraakt. „Ik zal natuurlijk


voldoen aan uw verzoek, maar ik zou wel eens willen weten, wat
daartoe de aanleiding is.”

„Meneer Geerling,” zei ze van het hooge standpunt eener dame, prat
op haar ongerepte deugd, tegenover de vleesch geworden
zedeloosheid, „het staat u dubbel leelijk den onschuldige [123]te
spelen. Overigens kunt u niet verwachten, dat iemand als ik over
zulke verachtelijke dingen met u spreek.”

Zij ging met meer drift, dan ze gewoonlijk aan den dag legde, het
huis binnen en liet hem staan.

Hij stond nog een oogenblik verbluft te kijken. Zoo iets had hij nooit
van haar verwacht. Welk een bejegening! En hoe volkomen
onverdiend! Hij was altijd beleefd en voorkomend tegen haar
geweest; had haar zijn kostgeld soms maanden vooruitbetaald,
wanneer zij c o u r t d ’ a r g e n t was; had nu zelfs getracht haar
trouweloozen man terug te brengen van zijn omgang met mevrouw
Du Roy, en tot loon.… kreeg hij een standje en werd op hoogst
onhebbelijke wijze als het ware de deur uitgezet.

Doch hij wilde het er niet bij laten: den volgenden ochtend zou hij
haar om een nadere verklaring verzoeken, al begreep hij reeds heel
goed wat ze bedoelde. Natuurlijk had men haar verteld, dat hij,
Geerling, dat onschuldige lam van een Van Brakel op den „slechten
weg” had gebracht. ’t Was waarlijk om te lachen!

Zijn vooronderstelling was juist.

Lucie, niet wetend wat te doen, buiten staat een krachtig besluit te
nemen, en toch in geen geval lijdzaam de rol van publiek bedrogene
willende spelen, had zich, toen Van Brakel en Geerling waren
vertrokken, gekleed en was naar mevrouw Willert gereden, die
vreemd opkeek over het zeer onverwacht bezoek.

De bleekheid en zenuwachtigheid der jonge vrouw konden haar wel


doen raden, wat er aan de hand was; het deed haar pleizier, dat
Lucie het te weten was gekomen, en ook [124]dat Willert uit was naar
een vergadering, zoodat ze samen ongestoord en vertrouwelijk
konden praten.

„Wel, wat voert je zoo laat hierheen?” vroeg ze, als wist zij van den
prins geen kwaad.

„Ik kom u een verzoek doen.”

„Wat zeg je dat ernstig! Is het zulk een zaak van gewicht?”

„Ja, dat is het,” antwoordde Lucie met een bevende stem. „Ik wilde
.… ik had u .… gaarne willen verzoeken.…”

Mevrouw Willert zag welk een moeite Lucie had om met haar
verzoek voor den dag te komen, en hoe zij streed om de tranen te
bedwingen, die met geweld te voorschijn drongen. Ze had
medelijden met haar, en nam met iets als moederlijke teederheid
haar hand.

„Je wenscht, dat ik Willert zal vragen je man te doen overplaatsen,


nietwaar?”

„Ja,” zei Lucie, die voor een oogenblik geheel opging in


verwondering, „ja, hoe weet u dat?”

„Wel, ik heb ’t zelf al eens aan mijn man gevraagd; het is voor Van
Brakel hier niet goed op de plaats.”

„Maar weet u dan.…?” vroeg Lucie angstig.

„Zeker. Ik weet immers, evenals iedereen, dat je man haast elken


avond in de sociëteit zit te dobbelen.”

„Ja, dat is zoo.”

„Welnu, hij werkt zich er op die manier totaal onder.”


„Och, dat zag ik zoo niet in.… Bij mijn papa thuis werd ook veel
gespeeld met een hoog tarief.”

„Dat is heel wat anders. Bij ondernemers kan dat; hun heele bedrijf
is eigenlijk één loterij; goed of slecht weer [125]scheelt hun duizenden
guldens; hooge of lage prijzen eveneens. En te voren weten ze
volstrekt niet wat ze krijgen zullen, en ze kunnen er hoegenaamd
niets aan doen. Maar bij een ambtenaar met een vast maandelijksch
inkomen, is dat anders; die moet niet dobbelen.”

„Er is toch niets met hem gebeurd?”

„Neen, maar als hij zoo voortgaat, is het t’ avond of morgen


onvermijdelijk. Hij dobbelt en verliest, en is toch altijd in den laatsten
tijd ruim in middelen. Ik vind dat je groot gelijk hebt, hoor, door
overplaatsing te wenschen. Het is hier op de plaats volstrekt niet
goed voor je man.”

Een rilling voer Lucie door de leden; het werd haar vreeselijk
duidelijk: Herman liet zich betalen.… O, dat was wel het
schandelijkste van alles! Onder den druk dier schande boog ze het
hoofd. Zij kon het niet langer uithouden; ’t was te erg.

Mevrouw Willert liet haar eerst een poosje uithuilen, richtte daarna
haar hoofd op, veegde de tranen van haar gelaat en kuste haar.

„Kom,” zei ze met meer aandoening, dan ze had willen doen blijken,
„wees zoo verdrietig niet, kind. Hij is het waarachtig niet waard.”

„Het is ook niet over hem; ik vind het zoo’n verschrikkelijke


schande.”

„Ja, fraai is het zeker niet; maar er is voor ’t moment niets aan te
doen; dus houd je goed. Naderhand kan je ’t hem wel inpeperen.”
„Nimmer,” zei Lucie. „Ik wil niets meer met hem te maken hebben.”
[126]

Ze hadden elkaar niets verteld; de naam van mevrouw Du Roy was


niet over haar lippen gekomen, maar toch spraken ze er over in de
volle overtuiging, dat zij elkaar volkomen begrepen.

„Eigenlijk,” hernam mevrouw Willert, „is het nog zijn eenige excuus.
Ik dacht eerst, dat hij het had gedaan uit voorkeur, en daar begreep
ik niets van. Nu ik er meer van ben te weten gekomen, is het me
duidelijker geworden; het spel en de schulden hadden hem beet, en
die lieve jongen, die bij jullie in ’t paviljoentje woont, schijnt hem
gebracht te hebben, waar hij wezen moest.”

„Meneer Geerling?”

„Dezelfde!”

„God, God, hoe is ’t mogelijk?”

„E n f i n , kindlief, ga jij nu rustig naar huis. Ik zal er Willert straks


over spreken, en ’t zou me al zeer verwonderen, als je niet heel
spoedig verhuisdet naar een gezonder klimaat.”

De gelegenheid om Lucie nader te interpelleeren deed zich voor


Geerling niet dadelijk voor, en hij moest zelf erkennen dat hij er niet
hard naar zocht.

’s Morgens scheen de zaak hem in een ander licht, en gevoelde hij


zich niet zoo geheel onschuldig als ’s avonds te voren. Hij was
bijzonder vroeg op en te gelijk met Van Brakel aan het ontbijt, die in
’t nauwe éénpersoons-ledikant in de logeerkamer haast niet had
kunnen slapen, en dat juist terwijl hij voor zijn doen bijzonder vroeg
was thuis gekomen. [127]

Lucie deed als gewoonlijk haar huishouden, maar haar gezicht was
zoo bleek en stond zoo strak, dat Geerling volstrekt geen lust
gevoelde om met haar in p o u r p a r l e r s te treden.

„Dat is nu een mooie geschiedenis,” zei hij nijdig tegen Van Brakel,
toen zij naar achter was gegaan.

„Hoe dan?”

„Wel, je vrouw geeft mij van alles de schuld, en heeft me op


schrikkelijk onaangename manier zoo goed als het huis ontzegd.”

Van Brakel keek hem aan met een suf gezicht, maar hij zei niets.

„Je begrijpt,” vervolgde Geerling op denzelfden toon, „dat ik er geen


gras over laat groeien. Ik pak mijn boeltje en trek straks naar het
logement. Als ik nu nog hier aan het ontbijt ben, dan is het alleen
omdat ik je het een en ander wilde meedeelen.”

„Och kom,” zei Van Brakel vergoelijkend: „je moet dat zóó niet
opvatten!”

„Die vind ik goed!” riep Geerling met hoog opgetrokken


wenkbrauwen en vol verbazing. „Die vind ik verdomd goed! Moet ik
dat zóó niet opvatten, zeg je! Wel, jij bent een leuke kerel!”

„Wel neen. Doe maar net of je niks hebt gehoord.”

„M e r c i a m i c e ! Dank je feestelijk, hoor! Ik heb er genoeg van, ik


ga er van door. Wees jij gerust je eigen bliksemafleider. Ik bedank
voor de eer en het genoegen.”
Van Brakel glimlachte pijnlijk en loosde een zucht. Het was waar. ’t
Speet hem zeer, dat Geerling nu heenging, [128]maar niet omdat hij
van ’t jonge mensch hield. Hij zag er tegen op nu met Lucie altijd
samen te zijn, en niets te hebben om een houding aan te nemen,
dan de kinderen, die erg verwend en lastig waren. Geerling was
altijd een geschikte figuur om tegen te praten en zoo, als men ’t
liever niet deed tegen elkaar.

„Het staat zoo gek,” zei hij nog. „Men zal er, God weet wat, van
zeggen.”

„Nu ja,” zei Geerling op een toon, die duidelijk te kennen gaf, dat
zulk een argument in den mond van iemand als Van Brakel geen de
minste waarde had.

„Het is niet om mij.”

„Zoo. Nu, mij kan het niet schelen. Ik zal Ceciel wel vertellen, wat er
van is.”

„Je schijnt dus nog steeds bij haar te komen?”

„Alsof je dat niet wist!”

„Neen, ik dacht dat je er mee waart geëindigd. Heb je werkelijk nog


altijd plan, die relatie voort te zetten?”

„Ik begrijp niet hoe je dat bedoelt,” zei Geerling koel en kort.

„Nu ja,” zei thans Van Brakel op zijn beurt.

Maar Geerling stoof woedend op. Hij verbood Van Brakel op die
manier te spreken over een fatsoenlijk meisje, dat het slachtoffer
was van een ongelukkige bloedverwantschap; dat paste h e m , die
daartoe op een schandelijke manier het zijne bijdroeg, wel ’t
allerminst. Wie iets zei ten nadeele van dat meisje, wier naam hij
zelfs niet in zulk een dispuut wilde brengen, die kreeg met hem te
doen, en die zou er onderuit moeten komen, zoo hij een man was.

Van Brakel in doodsangst, dat zijn vrouw het zou hooren, [129]trachtte
Geerling te kalmeeren, wat hem eerst met moeite en al zoete
broodjes bakkend gelukte. Maar inwendig was hij woedend. Zoo’n
aap, zoo’n uitgemergelde kwajongen, dien hij kon maken en breken,
stond daar tegen hem op te spelen en aan te gaan! O, als hij niet
bang was geweest voor Lucie, hij zou dat kereltje m o r e s hebben
geleerd.

Nog opgewonden ging Geerling naar zijn kamer, waar het weder
emballeeren zijner talrijke b i b e l o t s hem ruimschoots gelegenheid
gaf te bedaren. Toen hij klaar was, nam hij een bad, kleedde zich
keurig netjes en ging naar binnen, waar Lucie met haar naaister
linnengoed zat te herstellen.

„Ik wenschte u wel even alleen te spreken, mevrouw,” zei hij bedaard
en beleefd.

Zij zond de meid weg.

„U hebt me gisteravond aangezegd, dat ik zoo spoedig mogelijk een


goed heenkomen moest zoeken, en u ziet dat ik aan uw verlangen
dadelijk heb voldaan.”

„Dat is maar goed ook,” zei ze snibbig. „Hoe eer hoe beter.”

„Ik begrijp wat de reden is van de beleedigende manier, waarop mij


door u de deur wordt gewezen.”

„Ja, dàt geloof ik wel.”


„En ik ben alleen hier gekomen om u te zeggen, dat uw man reeds
lang bij mevrouw Du Roy aan huis kwam, voordat ik het wist, en dat
ik hem meermalen heb aangeraden, zijn omgang met die vrouw te
staken en er niet meer aan huis te komen.”

Zij stond op en glimlachte met minachting. [130]

„Juist, ik begrijp dat. ’t Was anders niet noodig mij met zulke
leugenachtige praatjes lastig te vallen. Ik ben te goed op de hoogte.”

„Mevrouw Van Brakel, ik geef u mijn woord van eer.…”

„Ik dank u, meneer Geerling; ik heb uw verzekeringen of


eerewoorden niet noodig; ik weet er alles van. Als u mij niets meer te
zeggen hebt, dan groet ik u, en ik hoop u nimmer weer te
ontmoeten.”

Zij was in zulk een opgewonden toestand geraakt, dat het


onmogelijk scheen haar tot een andere opinie te brengen.

Hij haalde de schouders op, maakte even een buiging en ging heen.

Het werden moeilijke dagen om te doorleven.

Lucie zag er slecht uit, al hield ze zich goed. Zij trachtte haar leed te
verzetten met werken en veel drukte en beweging te maken in huis.
Het gelukte haar een beetje. Maar somtijds viel ze op een stoel neer
in haar kamer en keek er versuft rond. Dan scheen het haar, alsof
het een droom was; alsof het niet mogelijk was, dat haar vreedzaam
en rustig huiselijk geluk in zulk een peilloos ravijn te pletter was
gevallen.

Van Brakel zag ze haast niet. Als ze geld moest hebben, liet ze ’t
hem vragen, en dan gaf hij het. Ze ontving het met een rilling van
afgrijzen. Wie weet hoe hij aan dat geld kwam! O, als zij geen
kinderen had gehad! Hij vermeed haar. ’s Morgens heel vroeg
ontbeet hij, en dan zorgde zij, dat alles klaar stond, maar zelve
verscheen ze niet in de achtergalerij; daar kwam ze eerst, als zij de
b e n d y hoorde wegrijden; ’s middags rijsttafelde zij met de kinderen
zeer [131]vroeg, en hij kwam eerst tegen twee uren uit de sociëteit
met een soms veel te zware lading bitter in de maag; hij dronk zijn
thee in de logeerkamer en zij achter. In den vooravond ging hij uit en
alleen ’s avonds te acht uren ontmoetten ze elkaar aan tafel.

Zij zag hem niet aan, en als hij een enkele maal trachtte een woord
in het midden te brengen, gaf zij geen antwoord; zoo bemerkte ze
ook niet, dat hij er hoe langer hoe gemeener begon uit te zien, dat
zijn gezicht meer opgeblazen was dan anders en zijn oogen steeds
rooder en gezwollener waren, door het dagelijksch misbruik van
spiritualiën, dat al ’s morgens begon en meestal eerst des nachts
eindigde. Hij bespeurde wel, dat ze vermagerde en bleeker was dan
ooit. En dat deed hem zielsleed, zóó dat hij meermalen zijn uiterste
best moest doen om niet te schreien. Hij was zichzelven niet meer,
dan alleen bij het spel. Een „fijntje” om een gulden het fiche met een
oploopenden pot of een poeterpartijtje, dan wel een partij biljart om
een ringit de carambole—dàt was met een brendy-soda, nog het
eenige, dat hem kon ontrukken aan den invloed van het beroerde
leven, en daarom zocht hij er meer dan ooit zijn troost bij.

Of daar iets kwaads in stak? Het idee kwam zelfs niet bij hem op.
Integendeel, hij vond het onder alle omstandigheden bijna het
eenige, wat een mensch had.

Lucie had een briefje gekregen van mevrouw Willert. „Het” was in
orde; de overplaatsing was telegraphisch bericht geworden; zij
moest maar dadelijk haar voorbereidende maatregelen nemen. Nu
het zoover was, zag ze er toch tegen op, en terwijl ze met den brief
in de hand voor zich uitstaarde, [132]drong al de s o e s a h en de
moeilijkheid dier verre overplaatsing naar een standplaats in het
binnenland, zich aan haar geest op; ze had daar eigenlijk geen lust
in; een gevoel van moedeloosheid en melancholie overviel haar, en
zuchtend keek ze droomerig naar de haar omringende omgeving;
die was niet opwekkend; de ondergaande zon was achter een dikke
laag loodgrijze wolken gezonken, die rustig aan den horizon lag en
waartegen de kampong aan de overzijde van den weg grauw en
prozaïsch zijn breedgebladerde pisangboomen afteekende; maar uit
den dikken wolkenrand schoot een breede roode lichtstroom, die in
het zenith tot het teederste rose vervloeide, en waarvan de
weerschijn in het zuiden en noorden vreemde lauwe kleuren
tooverde, purper en violetachtig; de smalle sikkel der opkomende
maan kwam in die kleurenvermenging uit met een ziekelijk
bleekgroen tintje. En dit melancholische namiddagtafereel, snel
wisselend met afnemende lichtsterkte, werkte niet mede om de arme
Lucie de toekomst met meer hoop en opgewektheid te gemoet te
doen zien.

Met een diepen zucht trachtte zij die neerslachtigheid af te


schudden. Komaan, het m o e s t ; het was ’t beste. En toen ze zich
dat eenmaal goed in het hoofd had geprent, zette zij haar bedienden
aan het werk, en liet de kisten en koffers voor den dag halen, die
met stof en spinrag overdekt in de bijgebouwen in een rommelkamer
stonden.

Het viel Van Brakel op, dien avond, dat er leege kisten stonden bij de
achtergalerij. Wat zou er nog meer gebeuren? Zou zij heengaan en
hem in den steek laten? Maar dàt zou hij in geen geval toestaan.
Doch wat zou hij vragen? [133]

Onder het eten verwonderde het haar zeer, dat hij niets zei; ze
verkeerde in de vaste vooronderstelling, dat zijn overplaatsing hem

You might also like