Research On Image Edge Detection Using The Genetic Algorithm

The image edge detection is the key part of image processing. Classical edge detections always have different localization. The task of finding the edges in an image that correspond to true physical boundaries remains a hotspot of image processing. Genetic Algorithm or GA is a kind of random search algorithm which uses natural selection and natural inheritance mechanism for reference. It was put forward by J.Holland in 1975. Genetic Algorithm is a kind of entirely optimizing search algorithm. It is provided with many characteristics, such as simpleness, currency, robustness, and fit for parallel processing. People did a lot of researches in how to use genetic algorithm to the image edge detection, but most of them are based on classical edge detection.This paper, based on classical edge detection and the comparative cost function, applies GA to the image edge detection using a kind of new coding scheme. And then, it brings forward a new hybrid genetic algorithm combining the excellences of the genetic algorithm and the tabu search, and applies this kind of new algorithm to image edge detection. The results of experiment indicate that the new algorithm has fine effect, high stability and excellent anti-noise capability.The main investigative content of this article are as follows:Firstly after introducing and analyzing the theory of classical edge detection, it points out the advantages and disadvantages, improving measure of each detection algorithm. Then the effects of several classical edge detection algorithms are put out and their simulation results are compared each other.Secondly aiming at classical algorithm, this paper introduces the image edge detection technique that is based on comparing cost function particularly and then presents a kind of improving iterative search algorithm to realize the image edge detection. It also analyses the iterative search algorithm and the experimental results, and compares the experimental results with classical edge detection. The new method performs very well.Thirdly based on the comparative cost function approach, the standard genetic algorithm is used to optimize the cost function and then the edge in an image is detected. In this paper, a kind of new chromosome coding scheme is presented, that is, taking two-dimension Boolean matrix as the chromosome, every chromosome represents a possible edge image. This kind of coding scheme can calculate the fitness of each individual without decoding, and the results of experiment show that the satisfactory edge images could be detected.Finally in allusion to the characteristics of standard genetic algorithm of premature convergence, a new hybrid algorithm that combines tabu search and genetic algorithm is investigated and it overcomes the phenomenon of standard genetic algorithm such as premature convergence. The new algorithm has characteristics of fast convergence speed and strong robustness. At last, the new algorithm is applied to different types of image; the experimental results show that it could detect lots of different types of image. The effects of edge detection are improved evidently. Moreover, it is provided with anti-noise strongly.