Mixde Color Image Segmentation Algorithm Research

Image segmentation is the process that classifies images with the same characteristics to the different attributes and extracts interested target,which is one of research focus question in the field of image understanding and machine.Its relevant segmentation algorithm emerges in endlessly,and is related to pattern classification,fuzzy mathematics,biological evolution,machine learning,graph theory and other interdisciplinary theory.In early year,people has studied gray image in dept.Along with the advance of computer technology the application of color image is increasing in daily life,and thus image segmentation also attracts attention.So this paper makes some improvement on the basis of the previous research.This paper mainly focused on studying natural color image segmentation algorithm and introduced the basic knowledge of digital image,region growing method,the theory of wavelet transform and so on.The main research results of this paper include the following two aspects:Firstly,we do some following improvements for regional segmentation algorithm based on ACO:1.By using intelligent characteristics of ant colony algorithm,adaptive selection of the initial seed points,the self-adaptive selection of the initial seed points can be realized,then according to the experiment to determine the threshold of establishing more accurate background model,effectively improve the segmentation accuracy and adaptability to the complex background changes.2.In view of the region growing and fusion rule,this paper used the relative Euclidean distance as a measure of region growing and fusion to do some analysis and discussion,and then got reasonable threshold value through experiments.3.Doing subsequent optimization processing by using the mathematical morphology method for the results of segmentation success can avoided effectively the rough problem of edge resulting from the external causes,such as noise problem,and improved the accuracy of the segmentation.The experiments proved that effectiveness of improved SRG segmentation algorithm segmentation is better.Secondly,for new image segmentation algorithm after combining neutral set Gabor transformation,this paper made the following several aspects:1.In view of the fixed shortcoming of operator ? and ? in the NS segmentation algorithm,this paper put forward a adaptive operator ? to substitute ?in order to improve the adaptability of algorithm.2.By using texture features of the Gabor wavelet transform to extract texture features of images,this paper put forward neutral set segmentation algorithm which combined color and texture feature to enhance the robustness of the algorithm to noise and improve the operation efficiency of the algorithm.The experimental results showed that segmentation effectiveness for the natural image after using improved algorithm is better.