In this paper an approach is described for segmenting medical images. We use active contour model, also known as snakes, and we propose an energy minimization procedure based on Genetic Algorithms (GA). The widely recognized power of deformable models stems from their ability to segment anatomic structures by exploiting constraints derived from the image data together with a priori knowledge about the location, size, and shape of these structures. The application of snakes to extract region of interest is, however, not without limitations. As is well known, there may be a number of problems associated with this approach such as initialization, existence of multiple minima, and the selection of elasticity parameters. We propose the use of GA to overcome these limits. GAs offer a global search procedure that has shown its robustness in many tasks, and they are not limited by restrictive assumptions as derivatives of the goal function. GAs operate on a coding of the parameters (the positions of the snake) and their fitness function is the total snake energy. We employ a modified version of the image energy which consider both the magnitude and the direction of the gradient and the Laplacian of Gaussian. Experimental results on synthetic images as well as on medical images are reported. Images used in this work are ocular fundus images, snakes result very useful in the segmentation of the Foveal Avascular Zone (FAZ). The experiments performed with ocular fundus images show that the proposed method is promising in the early detection of the diabetic retinopathy.
The question of how sensory channels may be autonomously constructed using generation and selection. The context is the discrimination of geometric shapes. In a first experiment, elements of a solution were attributed fitness based on the part of the problem they solved. In two subsequent experiments, cooperation between elements was respectively required and encouraged by means of a fitness function which only rewards complete solutions. Differences between the approaches are discussed, and generation and selection is concluded to provide a successful mechanism for the autonomous construction of sensory channels.
Image processing is usually done by chaining a series of well known image processing operators. Using evolutionary methods this process may be automated. In this paper we address the problem of evolving task specific image processing operators. In general, the quality of the operator depends on the task and the current environment. Using genetic programming we evolved an interest operator which is used to calculate sparse optical flow. To evolve the interest operator we define a series of criteria which need to be optimized. The different criteria are combined into an overall fitness function. Finally, we present experimental results on the evolution of the interest operator.
A technique is described for the optimisation of spatio-temporal (3-D) grey-scale soft morphological filters for applications in archive film restoration. The optimisation is undertaken using genetic algorithms. By employing filters which incorporate the temporal dimension, this technique extends and improves upon previously described techniques which were based purely in the spatial (2-D) domain. Examples of applying the technique to real-world film restoration problems are shown.
The long term goal of the work described in this paper is the development of a bio-inspired system, employing evolvable hardware, that adapts according to the needs of the environment in which it is deployed. The application described here is the design of a novel and highly parallel image processing tool to detect edges within a wide range of conventional grey-scale images. We discuss the simulation of such a system based on a genetic programming paradigm, using a simple binary logic tree to implement the genetic string coding. The results acquired from the simulation are compared with those obtained from the application of a conventional Sobel edge detector, and although rudimentary, show the great potential of such bio-inspired systems.
The traditional paradigm for digital filter design is based on the concept of a linear difference equation with the output response being a weighted sum of signal samples with usually floating point coefficients. Unfortunately such a model is necessarily expensive in terms of hardware as it requires many large bit additions and multiplications. In this paper it is shown how it is possible to evolve a small rectangular array of logic gates to perform low pass FIR filtering. The circuit is evolved by assessing its response to digitised pure sine waves. The evolved circuit is demonstrated to possess nearly linear properties, which means that it is capable of filtering composite signals which it has never seen before.
In this work we examine the applicability of an evolutionary strategy to the problem of fitting constrained second-order surfaces to both synthetic and acquired 3D data. In particular we concentrate on the Genocop III algorithm proposed by Michalewicz for the optimization of constrained functions. This is a novel application of this algorithm which has demonstrably good results when applied using parametric models. Example times for convergence are given which compare the approach to standard techniques.
This paper describes a technique for model-based object recognition in a noisy and cluttered environment, by extending the work presented in an earlier study by the authors. In order to accurately model small irregularly shaped objects, the model and the image are represented by their binary edge maps, rather then approximating them with straight line segments. The problem is then formulated as that of finding the best describing match between a hypothesized object and the image. A special form of template matching is used to deal with the noisy environment, where the templates are generated on-line by a Genetic Algorithm. For experiments, two complex test images have been considered and the results when compared with standard techniques indicate the scope for further research in this direction.
A technique is proposed to reduce the peak power consumption of sequential circuits during test pattern application. High-speed computation intensive VLSI systems, as telecommunication systems, make power management during test a critical problem. A Genetic Algorithm computes a set of redundant test sequences, then a genetic optimization algorithm selects the optimal subset of sequences able to reduce the consumed power, without reducing the fault coverage. Experimental results gathered on benchmark circuits show that our approach decreases the peak power consumption by 20% on the average with respect to the original test sequence generated ignoring the power dissipation problem, without affecting the fault coverage.
In this paper we investigate the application of a combined Genetic Programming - Simulated Annealing (GP-SA) solution to a classical signal processing problem which arises in communications systems. This is the so called channel equalisation problem where the task is to construct a system which adaptively compenstates for imperfections in the path from the transmitter to the receiver. Our primary interest is to examine the recosntruction of binary data sequences transmitted through distorting channels. We measure the performance of a generic GP-SA equaliser and compare it to that of standard methods commonly used in real systems. In particular, we consider special cases which are known to be difficult for the existing methods, such as non-linear and partial response channels. Our results show that the GP-SA method generally offers superior signal restoration but at the expense of computational effort.
This paper introduces a new method of performing mutation in a real-coded Genetic Algorithm (GA), a method driven by Principal Component Analysis (PCA). We present both theoretical and empirical results which show that our mutation operator attains higher levels of diversity in the search space, as compared to other mutation operators, meaning that by employing our mutation operator we maintain diverse populations that increase the chances of finding better solutions during evolution of the GA. The performances of the real-coded GA with PCA-mutation were checked on the problem of designing IIR filters by Deczky method, which is a well known direct design method of IIR filters. Results obtained show that our PCA-mutation GA has been more successful in keeping diverse populations during search, the consequence being the finding of better solutions to the filter design problem, compared to solutions found using GA with classical mutation operators.
The paper addresses an important and difficult problem of object recognition in poorly constrained environments and with objects having large variability. This research uses genetic programming (GP) to develop automatic object detectors. The task is to detect vehicles in infrared line scan (IRLS) images gathered by low flying aircraft. This is a difficult task due to the diversity of vehicles and the environments in which they can occur, and because images vary with numerous factors including fly-over, temporal and weather characteristics. A novel multi-stage approach is presented which addresses automatic feature detection, automatic object segregation, rotation invariance and generalisation across diverse objects whilst discriminating from a myriad of potential non-objects. The approach does not require imagery to be pre-processed.
Evoked potentials are electrical signals produced by the body in response to a stimulus. In general these signals are noisy with a low signal to noise ratio. In this paper a method is proposed that uses sets of filters, whose cut-off frequencies are selected by an evolutionary algorithm. An evolutionary algorithm was investigated to limit the assumptions that were made about the signals. The set of filters separately filter the evoked potentials, and are combined as a weighted sum of the filter outputs. The evolutionary algorithm also selects the weights. Inputs to the filters are sets of averaged signal, 4 or 10 signals per average. Even though there is likely to be variations between the signals, this process can improve the extraction of potentials.
This paper presents a new approach to fast 3D analysis of stereo images, based on an Evolutionary Strategy. The algorithm evolves a population of three-dimensional points in the camera field of view, so that best fitness values are obtained when the points lie on the surfaces of 3-D objects. In this Co-Evolutionary method, the solution of the problem is the whole population rather than a single individual. The algorithm uses a low-cost fitness function, sharing and simple mutation and crossover operators. Some results on test images are presented and potential extensions to image sequence processing, moving objects tracking and mobile robot obstacle avoidance are discussed.
This paper describes how sets of visual feature detectors are evolved starting from simple primitives. The primitives, of which some are inspired on visual processing observed in mammalian visual pathways, are combined using genetic programming to form a feed-forward feature-extraction hierarchy. Input to the feature detectors consists of a series of real-world images, containing objects or faces. The results show how each set of feature detectors self-organizes into a set which is capable of returning feature vectors for discriminating the input images. We discuss the influence of different settings on the evolution of the feature detectors and explain some phenomena.