Cristian Munteanu

Gray-scale image enhancement as an automatic process driven by evolution

Image enhancement is the task of applying certain transformations to an input image such as to obtain a visually more pleasant, more detailed, or less noisy output image. The transformation usually requires interpretation and feedback from a human evaluator of the output result image. Therefore, image enhancement is considered a difficult task when attempting to automate the analysis process and eliminate the human intervention. This paper introduces a new automatic image enhancement technique driven by an evolutionary optimization process. We propose a novel objective criterion for enhancement, and attempt finding the best image according to the respective criterion. Due to the high complexity of the enhancement criterion proposed, we employ an evolutionary algorithm (EA) as a global search strategy for the best enhancement. We compared our method with other automatic enhancement techniques, like contrast stretching and histogram equalization. Results obtained, both in terms of subjective and objective evaluation, show the superiority of our method.

Using assortative mating in genetic algorithms for vector quantization problems

In nature, some species mate according to their phenotype similarity. The Assortative Mating Genetic Algorithm (AMGA) mimics some mechanisms of reproduction in natural environments. The main difference between AMGA and the Standard GA (SGA) is the selection of the parents in the crossover operators. We develop a similarity measure for the Vector Quantization problem and we show that the application of AMGA to some instances of this problem reduces the number of times that the algorithm becomes trapped in local optima. We also present results that show that AMGA keeps a higher level of genetic diversity than the SGA.

Evolutionary image enhancement with user behavior modeling

In this paper we present a novel method for image enhancement of gray-scale images based on the simulation of evolution. Our method employs Genetic Algorithms to evolve the shape of the contrast curve in the image, while attempting to partially automate the subjective process of image evaluation (e.g. user behavior) by performing multiple regression on fitness values. Results obtained show the robustness and efficiency of the evolutive method for image enhancement. For several images in the test set our method obtains better results than the classical histogram equalization technique. Extensive statistics performed, shows that multiple regression can be effectively applied to model the user behavior.

Asymmetric hemisphere modeling in an offline brain-computer interface

Classification of the electroencephalogram (EEG) during motor imagery of the left or right hand can be performed using a classifier comprising two hidden Markov models (HMMs) describing the spatio-temporal patterns related to the imagination. Due to the known asymmetries during motor imagery of rightand left-hand movement, an HMM-based classifier allowing asymmetrical structures is introduced. The comparison between such a system and a symmetrical one is based on the error rate of classification. The results for EEG data collected during 20 sessions from five subjects demonstrate a significant improvement of 9% for the classification accuracy for the asymmetric classifiers. The selection of the DAM for classification is done using a variant of genetic algorithms (GAs); namely, the adaptive reservoir genetic algorithm (ARGA).

CAP event detection by wavelets and GA tuning

The electroencephalogram (EEG) is a non-invasive and affordable technique to study the brain activity during wakefulness and especially during sleep. The current trend on sleep analysis focuses on its dynamic organization described in the cyclic alternating pattern (CAP) paradigm. This paradigm looks to the EEG microstructure, gives attention to the short duration EEG phenomena instead of the traditional global view of fixed epochs of 20 or 30 seconds, heralded by the Rechtschaffen and Kales (R&K) classification. The CAP paradigm not only cares about the short duration events but also models its periodic activity, providing thus, important information on EEG synchrony modulation in the sleep process. These periodic activities are characterized by repeated spontaneous and or evoked EEG activations or transient responses (A phases), at intervals up to one minute emerged from the background activity (B phases). In a previous work we proposed an automatic detector and classifier of A phases in sleep EEG, where wavelet transforms is used to analyze the sleep EEG signal in the time-frequency domain, in order to separate the signal power in different and physiologically relevant frequency bands. The CAP A phases characteristics described in the classification Atlas is used in the design of the structure and the starting set of parameters of the detector and classifier. The objective of this work is to use genetic algorithm to tune the parameters of the detector and classifier. Excerpts of all night sleep recordings of a group of volunteers are used in the tuning and testing. Results are compared with visual and other automatic classification.

Speckle Reduction Through Interactive Evolution of a General Order Statistics Filter for Clinical Ultrasound Imaging

This communication presents an interactive tool performing adaptive speckle filtering so that the medical expert who runs the algorithm has permanent control over the output and guides the process towards obtaining enhanced images that agree to his/her subjective quality criteria. The core of the filtering tool is an interactive genetic algorithm that adapts online the coefficients of a general order statistics filter. Preliminary results show the potential of the method in comparison to other powerful speckle reduction filters on a test bed comprising obstetrics and gynecology ultrasound images.

Evolutionary image enhancement with user behaviour modeling

paper we present a novel method for image enhancement of gray-scale images based on the simulation of evolution. Our method employs Genetic Algorithms to evolve the shape of the contrast curve in the image, while attempting to partially automate the subjective process of image evaluation (e.g. user behaviour) by performing multiple regression on fitness values. Results obtained show the robustness and efficiency of the evolutive method for image enhancement. For several images in th e test set our method obtains better results than the classical histogra equalization technique. Extensive statistics performed, show that multiple regression can be effectively applied to model the user behaviour.

Symmetry at the genotypic level and the simple inversion operator

Classical Genetic Algorithm theory was built on four operators: proportional selection, one-point crossover, mutation and inversion. While the role of inversion was questioned, the use of the other remaining operators has thrived, some of these newly designed operators being motivated by good empirical results, some having a solid theory to support their use. In this paper we present a Simple Inversion Operator, and we investigate its potential mixing capabilities for problems where the optimum consists of juxtaposed Symmetric Building Blocks. Both theoretical investigation and experimental results obtained, indicate that our operator is quite powerful in finding the right building blocks that compose the optimum, whenever symmetrical building blocks play an important role in the discovery of the global solution.

Adaptive Reservoir Evolutionary Algorithm: An Evolutionary On-Line Adaptation Scheme for Global Function Optimization

This paper introduces a novel global optimization heuristic algorithm based on the basic paradigms of Evolutionary Algorithms (EA). The algorithm greatly extends a previous strategy proposed by the authors in Munteanu and Lazarescu (1998). In the newly designed algorithm the exploration/exploitation of the search space is adapted on-line based on the current features of the landscape that is being searched. The on-line adaptation mechanism involves a decision process as to whether more exploitation or exploration is needed depending on the current progress of the algorithm and on the current estimated potential of discovering better solutions. The convergence with probability 1 in finite time and discrete space is analyzed, as well as an extensive comparison with other evolutionary optimization heuristics is performed on a set of test functions.

Evolving Order Statistics Filters for Image Enhancement

This paper describes an effective method for performing both image denoising and contrast/brightness enhancement to images corrupted with a wide category of noise. The method employs a Real-Coded Genetic Algorithm with subjective fitness and a novel crossover operator called Gaussian Uniform Crossover. The algorithm evolves the structure of a general Order Statistics Filter (OSF). Results are presented that indicate the efficiency of the method proposed as compared to classical filtering methods.