A Research And Application Of Image Segmentation Based On Improved Genetic Algorithm

Image segmentation is a key basis of many higher level image processing activities such as visualization, compression, and image guided medical diagnoses. Numerous algorithms using different approaches have been proposed for image segmentation. Thresholding is a popular tool for image segmentation for its simplicity, especially in the fields where the real time processing is needed. However, its time-consuming computation is often an obstacle in real time application systems. So it is meaningful to develop an effective algorithm to solve the problem of image segmentation based on thresholding.Genetic Algorithm (GA) is a sort of efficient, parallel, global search method with its inherent virtues of robustness parallel and self-adaptive characters. It is suitable for searching the optimization result in the large search space. Now it has been applied widely and perfectly in many study fields and engineering areas. In computer vision field GA is increasingly attached more importance. It provides the image segmentation a new and effective method.In order to automatically determine the optimal threshold in image segmentation, a new method of image segmentation based on improved genetic algorithm is presented in this thesis, it is to use this improved Genetic Algorithm to globally optimize 2-mension OTSU image segmentation functions. This method can automatically adjust the parameters of Genetic Algorithm according to the fitness values of individuals and the decentralizing degree of individuals of the population, and keep the variety of population while rapidly converging to get the optimal thresholds in image segmentation, it overcomes the shortcomings including worse convergent speed, easy to be premature that exist in traditional Genetic Algorithm etc. The theoretical analysis and simulating experiments show that the range of the thresholds is more stable and it consumes less time greatly and better satisfies the request of real-time processing in image segmentation by using this new method, compared with 2-mension OTSU image segmentation and genetic algorithm based image segmentation.The innovations and main contents of the thesis can be concluded as follows:First, a new method of image segmentation based on improved Genetic Algorithm is proposed, which can optimize solution, particularly in the option of adaptive mutation operator that considered the characteristics of Genetic Algorithm and actual operation efficiency of the algorithm . Experiments proved that the improved algorithm for a noise disturbance had better grayscale image segmentation quality and markedly improved the running time compared with the traditional segmentation method, while taking advantage of improving the program.Second, an improved method of OTSU is proposed, in which a new distance measurement is proposed, that is, the distance between the background and objectives ,the two types of the greater distance between the target and background on sub-the more open the better segmentation results. In the improved OTSU, a new measure of good and bad cohesion variables is introduced, which is the average variance of the background and objectives, thereby introducing the concept of two types of the average variance is used to measure the cohesion, the smaller the average variance of two types have, the more uniform pixels each class has, the better cohesion and segmentation effect will be greater.Third, a new method of image segmentation based on improved Genetic Algorithm and improved OTSU is proposed.Fourthly, according to the simulation experiments, the improved AGA can preserve the multifamily of population and the astringency of the algorithm, threshold computing time is 18 ms (about 63%) shorter than the 2-mension OTUS image segmentation method, and reduces about 30% compared with the basic Genetic Algorithm; the stability of the global convergence of the algorithm is improved, and the threshold is less than 3 pixels range. The improved algorithm can quickly and accurately segment image, which can be used in a variety of real-time image processing and analysis with high practicality.