EvoWorkshops2005: EvoIASP

7th European Workshop on Evolutionary Computation in Image Analysis and Signal Processing

In Applications of Evolutionary Computing.

Evolutionary algorithms have been shown to be tools which can be used effectively in the development of systems (software or hardware) for image analysis and signal processing in complex domains of high industrial and social relevance.

EvoIASP has been the first European event specifically dedicated to the applications of evolutionary computation (EC) to image analysis and signal processing (IASP) and gives European and non-European researchers in those fields, as well as people from industry, an opportunity to present their latest research and to discuss current developments and applications, besides fostering closer future interaction between members of the three scientific communities.

Web address: http://www.evonet.info/eurogp2005

Topics include

Organising Committee

Program Chair
Stefano Cagnoni
cagnoni AT ce DOT unipr DOT it
University of Parma, Italy
 
EvoWorkshops2005 Chair
Franz Rothlauf
rothlauf AT uni-mannheim DOT de
University of Mannheim, Germany
 
Local Chair
Marco Tomassini
Marco.Tomassini AT hec DOT unil DOT ch
University of Lausanne, Switzerland
 
Publicity Chair
Jano van Hemert
jvhemert AT cwi DOT nl
Napier University, Edinburgh, Scotland, UK

Note: the e-mail addresses are masked for spam protection.

Programme Committee

Giovanni Adorni, University of Genoa (Italy)
Lucia Ballerini (Sweden)
Bir Bhanu, University of California, Riverside (USA)
Dario Bianchi, University of Parma (Italy)
Leonardo Bocchi, University of Florence (Italy)
Alberto Broggi, University of Parma (Italy)
Stefano Cagnoni, University of Parma (Italy)
Ela Claridge, University of Birmingham (UK)
Ernesto Costa, University of Coimbra (Portugal)
Laura Dipietro, Massachusetts Institute of Technology (USA)
Marc Ebner, University of Wuerzburg (Germany)
Terry Fogarty, South Bank University (UK)
Daniel Howard, Qinetiq (UK)
Mario Koeppen, FhG IPK Berlin (Germany)
Evelyne Lutton, INRIA (France)
Gustavo Olague, CICESE (Mexico)
Riccardo Poli, University of Essex (UK)
Conor Ryan, University of Limerick (Ireland)
Stephen Smith University of York (UK)
Giovanni Squillero, Politecnico di Torino (Italy)
Wolfgang Stolzmann, Daimler-Chrysler (Germany)
Kiyoshi Tanaka, Shinshu University (Japan)
Ankur M. Teredesai, Rochester Institute of Technology (USA)
Andy Tyrrell, University of York (UK)
Robert Vanyi, Siemens PSE (Hungary)
Hans-Michael Voigt, Center for Applied Computer Science (GFaI) (Germany)
Stewart Wilson, Prediction Dynamics (USA)
Mengjie Zhang, Victoria University of Wellington (New Zealand)


EvoIASP Programme

Wednesday, 30 March 2005

Session 1: Computer Vision Applications (1120-1250)

Evolving Parameters of Surveillance Video Systems for Non-Overfitted Learning
Óscar Pérez
Jesús García
Antonio Berlanga
José Manuel Molina

Automated Photogrammetric Network Design using the Parisian Approach
Enrique Dunn
Gustavo Olague
Evelyne Lutton
(Best Paper Award Candidate)

Sensory-motor Coordination in Gaze Control
Guido C.H.E. de Croon
Eric O. Postma
Jaap van den Herik
(Best Paper Award Candidate)

Session 2: Genetic Programming (1415-1545)

Learning Weights in Genetic Programs Using Gradient Descent for Object Recognition
Mengjie Zhang
Will Smart
(Best Paper Award Candidate)

A Multistage Approach To Cooperatively Coevolving Feature Construction and Object Detection
Mark Roberts
Ela Claridge

An Implicit Context Representation for Evolving Image Processing Filters
Stephen L. Smith
Stefan Leggett
Andrew M. Tyrrell

Session 3: Miscellaneous (1600-1730)

Genetic-Fuzzy Optimization Algorithm for Adaptive Learning of Human Vocalization in Robotics
Enzo Mumolo
Massimiliano Nolich
Graziano Scalamera

Design of Fast Multidimensional Filters Using Genetic Algorithms
Max Langer
Björn Svensson
Anders Brun
Mats Andersson
Hans Knutsson

Region Merging for Severe Oversegmented Images using a Hierarchical Social Metaheuristic
Abraham Duarte
Ángel Sánchez
Felipe Fernández
Antonio Sanz

Session 4: Posters (1730-2000)

A Coevolutionary Approach for Clustering with Feature Weighting Application to Image Analysis
Alexandre Blansché
Pierre Gançarski
Jerzy J. Korczak

A new evolutionary algorithm for image segmentation
Leonardo Bocchi
Lucia Ballerini
Signe Hässler

Object Detection for Computer Vision Using a Robust Genetic Algorithm
Tania Mezzadri Centeno
Heitor Silvério Lopes
Marcelo Kleber Felisberto
Lúcia Valéria Ramos de Arruda

An Interactive EA for Multifractal Bayesian Denoising
Evelyne Lutton
Pierre Grenier
Jacques Levy Vehel

An Evolutionary Infection Algorithm for Dense Stereo Correspondence
Cynthia Beatriz Pérez
Gustavo Olague
Francisco Fernandez
Evelyne Lutton

Automatic Image Enhancement Driven by Evolution Based on Ridgelet Frame in the Presence of Noise
Tan Shan
Shuang Wang
Xiangrong Zhang
Licheng Jiao

Practical Evaluation of Efficient Fitness Functions for Binary Images
Róbert Ványi

Selective SVMs Ensemble Driven by Immune Clonal Algorithm
Xiangrong Zhang
Shuang Wang
Tan Shan
Licheng Jiao


EvoIASP: Titles and abstracts of accepted papers

Alexandre Blansché
Pierre Gançarski
Jerzy J. Korczak

A Coevolutionary Approach for Clustering with Feature Weighting Application to Image Analysis

This paper presents a new process for modular clustering of complex data, such as that used in remote sensing images. This method performs feature weighting in a wrapper approach. The proposed method combines several local specialists, each one extracting one cluster only and using different feature weights. A new clustering quality criterion, adapted to independant clusters, is defined. The weight learning is performed through a cooperative coevolution algorithm, where each species represents one of the clusters to be extracted.


Leonardo Bocchi
Lucia Ballerini
Signe Hässler

A new evolutionary algorithm for image segmentation

This paper describes a new evolutionary algorithm for image segmentation. The evolution involves the colonization of a bidimensional world by a number of populations. The individuals, belonging to different populations, compete to occupy all the available space and adapt to the local environmental characteristics of the world. We present experiments with synthetic images, where we show the efficiency of the proposed method and compare it to other segmentation algorithm, and an application to medical images. Reported results indicate that the segmentation of noise images is effectively improved. Moreover, the proposed method can be applied to a wide variety of images.


Tania Mezzadri Centeno
Heitor Silvério Lopes
Marcelo Kleber Felisberto
Lúcia Valéria Ramos de Arruda

Object Detection for Computer Vision Using a Robust Genetic Algorithm

This work is concerned with the development and implementation of an image pattern recognition approach to support computational vision systems when it is necessary to automatically check the presence of specific objects on a scene, and, besides, to describe their position, orientation and scale. The developed methodology involves the use of a genetic algorithm to find known 2D object views in the image. The proposed approach is fast and presented a robust performance in several test instances including multiobject scenes, with or without partial occlusion.


Guido C.H.E. de Croon
Eric O. Postma
H. Jaap van den Herik

Sensory-motor Coordination in Gaze Control

(Best Paper Award Candidate)
In the field of artificial intelligence, there is a considerable interest in the notion of sensory-motor coordination as an explanation for intelligent behaviour. However, there has been little research on sensory-motor coordination in tasks that go beyond low-level behavioural tasks. In this paper we show that sensory-motor coordination can also enhance performance on a high-level task: artificial gaze control for gender recognition in natural images. To investigate the advantage of sensory-motor coordination, we compare a non-situated model of gaze control (incapable of sensory-motor coordination) with a situated model of gaze control (capable of sensory-motor coordination). The non-situated model of gaze control shifts the gaze according to a fixed set of locations, optimised by an evolutionary algorithm. The situated model of gaze control determines gaze shifts on the basis of local inputs in a visual scene. An evolutionary algorithm optimises the modelŐs gaze control policy. From the experiments performed, we may conclude that sensory-motor coordination contributes to artificial gaze control for the high-level task of gender recognition in natural images: the situated model outperforms the non-situated model. The mechanism of sensory-motor coordination establishes dependencies between multiple actions and observations that are exploited to optimise categorisation performance.


Abraham Duarte
Ángel Sánchez
Felipe Fernández
Antonio Sanz

Region Merging for Severe Oversegmented Images using a Hierarchical Social Metaheuristic

This paper proposes a new evolutionary region merging method to improve segmentation quality result on oversegmented images. The initial segmented image is described by a modified Region Adjacency Graph model. In a second phase, this graph is successively partitioned in a hierarchical fashion into two subgraphs, corresponding to the two most significant components of the actual image, until a termination condition is met. This graph-partitioning task is solved as a variant of the min-cut problem (normalized cut) using a Hierarchical Social (HS) metaheuristic. We applied the proposed approach on different standard test images, with high-quality visual and objective segmentation results.


Enrique Dunn
Gustavo Olague
Evelyne Lutton

Automated Photogrammetric Network Design using the Parisian Approach

(Best Paper Award Candidate)
We present a novel camera network design methodology based on the Parisian approach to evolutionary computation. The problem is partitioned into a set of homogeneous elements, whose individual contribution to the problem solution can be evaluated separately. These elements are allocated in a population with the goal of creating a single solution by a process of aggregation. Thus, the goal of the evolutionary process is to generate individuals that jointly form better solutions. Under the proposed paradigm, aspects such as problem decomposition and representation, as well as local and global fitness integration need to be addressed. Experimental results illustrate significant improvements, in terms of solution quality and computational cost, when compared to canonical evolutionary approaches.


Max Langer
Björn Svensson
Anders Brun
Mats Andersson
Hans Knutsson

Design of Fast Multidimensional Filters Using Genetic Algorithms

A method for designing fast multidimensional filters using genetic algorithms is described. The filter is decomposed into component filters where coefficients can be sparsely scattered using filter networks. Placement of coefficients in the filters is done by genetic algorithms and the resulting filters are optimized using an alternating least squares approach. The method is tested on a 2-D quadrature filter and the method yields a higher quality filter in terms of weighted distortion compared to other efficient implementations that require the same ammount of computations to apply. The resulting filter also yields lower weighted distortion than the full implementation.


Evelyne Lutton
Pierre Grenier
Jacques Levy Vehel

An Interactive EA for Multifractal Bayesian Denoising

We present in this paper a multifractal bayesian denoising technique based on an interactive EA. The multifractal denoising algorithm that serves as a basis for this technique is adapted to complex images and signals, and depends on a set of parameters. As the tuning of these parameters is a difficult task, highly dependent on psychovisual and subjective factors, we propose to use an interactive EA to drive this process. Comparative denoising results are presented with automatic and interactive EA optimisation. The proposed technique yield efficient denoising in many cases, comparable to classical denoising techniques. The versatility of the interactive implementation is however a major advantage to handle difficult images like IR or medical images.


Enzo Mumolo
Massimiliano Nolich
Graziano Scalamera

Genetic-Fuzzy Optimization Algorithm for Adaptive Learning of Human Vocalization in Robotics

We present a computational model of human vocalization which aims at learning the articulatory mechanisms which produce spoken phonemes. It uses a set of fuzzy rules and genetic optimization. The former represents the relationships between places of articulations and speech acoustic parameters, while the latter computes the degrees of membership of the places of articulation. That is, the places of articulation are considered as fuzzy sets whose degrees of membership are the articulatory features. Subjective listening tests of sentences artificially generated from the articulatory description resulted in an average phonetic accuracy of about 76Through the analysis of a large amount of natural speech, the algorithm can be used to learn the places of articulation of all phonemes.


Óscar Pérez
Jesús García
Antonio Berlanga
José Manuel Molina

Evolving Parameters of Surveillance Video Systems for Non-Overfitted Learning

This paper presents an automated method based on Evolution Strategies (ES) for optimizing the parameters regulating video-based tracking systems. It does not make assumptions about the type of tracking system used. The paper proposes an evaluation metric to assess system performance. The illustration of the method is carried out using three very different video sequences in which the evaluation function assesses trajectories of airplanes, cars or baggage-trucks in an airport surveillance application. Firstly, the optimization is carried out by adjusting to individual trajectories. Secondly, the generalization problem (the search for appropriate solutions to general situations avoiding overfitting) is approached considering combinations of trajectories to take into account in the ES optimization. In both cases, the trained system is tested with the rest of trajectories. Our experiments show how, besides an automatic and reliable adjustment of parameters, the optimization strategy of combining trajectories improves the generalization capability of the training system.


Cynthia Beatriz Pérez
Gustavo Olague
Francisco Fernandez
Evelyne Lutton

An Evolutionary Infection Algorithm for Dense Stereo Correspondence

This work presents an evolutionary approach to improve the infection algorithm to solve the problem of dense stereo matching. Dense stereo matching is used for 3D reconstruction in stereo vision in order to achieve fine texture detail about a scene. The algorithm presented in this paper incorporates two different epidemic automata applied to the correspondence of two images. These two epidemic automata provide two different behaviours which construct a different matching. Our aim is to provide with a new strategy inspired on evolutionary computation, which combines the behaviours of both automata into a single correspondence process. The new algorithm will decide which epidemic automata to use based on inheritance and mutation, as well as the attributes, texture and geometry, of the input images. Finally, we show experiments in a real stereo pair to show how the new algorithm works.


Mark Roberts
Ela Claridge

A Multistage Approach To Cooperatively Coevolving Feature Construction and Object Detection

In previous work, we showed how cooperative coevolution could be used to evolve both the feature construction stage and the classification stage of an object detection algorithm. Evolving both stages simultaneously allows highly accurate solutions to be created while needing only a fraction of the number of features extracting as in generic approaches. Scalability issues in the previous system have motivated the introduction of a multi-stage approach which has been shown in the literature to provide large reductions in computational requirements. In this work we show how using the idea of coevolutionary feature extraction in conjunction with this multi-stage approach can reduce the computational requirements by at least two orders of magnitude, allowing the impressive performance gains of this technique to be readily applied to many real world problems.


Stephen L. Smith
Stefan Leggett
Andrew M. Tyrrell

An Implicit Context Representation for Evolving Image Processing Filters

This paper describes the implementation of a representation for Cartesian Genetic Programming (CGP) in which the specific location of genes within the chromosome has no direct or indirect influence on the phenotype. The mapping between the genotype and phenotype is determined by self-organised binding of the genes, inspired by enzyme biology. This representation has been applied to a version of CGP developed especially for evolution of image processing filters and preliminary results show it outperforms the standard representation in some configurations.


Tan Shan
Shuang Wang
Xiangrong Zhang
Licheng Jiao

Automatic Image Enhancement Driven by Evolution Based on Ridgelet Frame in the Presence of Noise

Many conventional and well-known image enhancement methods suffer from a tendency to increase the visibility of noise when they enhance the underlying details. In this paper, a new kind of image analysis tool - ridgelet frame - is introduced into the arena of image enhancement. We design an enhancement operator with the advantages that it not only enhance image details but also avoid the amplification of noise within source image. Different from those published previously, our operator has more parameters, which results in more flexibility for different category images. Based on an objective criterion, we search the optimal parameters for each special image using Immune Clone Algorithm (ICA). Experimental results show the superiority of our method in terms of both subjective and objective evaluation.


Róbert Ványi

Practical Evaluation of Efficient Fitness Functions for Binary Images

Genetic Programming can be used to evolve complex objects. One field, where GP may be used is image analysis. There are several works using evolutionary methods to process, analyze or classify images. All these procedures need an appropriate fitness function, that is a similarity measure. However, computing such measures usually needs a lot of computational time. To solve this problem, the notion of efficiently computable fitness functions was introduced, and their theory was already examined in detail. In contrast to that work, in this paper the practical aspects of these fitness functions are discussed.


Xiangrong Zhang
Shuang Wang
Tan Shan
Licheng Jiao

Selective SVMs Ensemble Driven by Immune Clonal Algorithm

A selective ensemble of support vector machines (SVMs) based on immune clonal algorithm (ICA) is proposed for the case of classification. ICA, a new intelligent computation method simulating the natural immune system, characterized by rapid convergence to global optimal solutions, is employed to select a suitable subset of the trained component SVMs to make up of an ensemble with high generalization performance. The experimental results on some popular datasets from UCI database show that the selective SVMs ensemble outperforms a single SVM and traditional ensemble method that ensemble all the trained component SVMs.


Mengjie Zhang
Will Smart

Learning Weights in Genetic Programs Using Gradient Descent for Object Recognition

(Best Paper Award Candidate)
This paper describes an approach to the use of gradient descent search in tree based genetic programming for object recognition problems. A weight parameter is introduced to each link between two nodes in a program tree. The weight is defined as a floating point number and determines the degree of contribution of the sub-program tree under the link with the weight. Changing a weight corresponds to changing the effect of the sub-program tree. The weight changes are learnt by gradient descent search at a particular generation. The programs are evolved and learned by both the genetic beam search and the gradient descent search. This approach is examined and compared with the basic genetic programming approach without gradient descent on three object classification problems of varying difficulty. The results suggest that the new approach works well on these problems.