Research And Implementation Of Color Image Segmentation Algorithm

Image Processing has far-ranging application in our life, and the most attentions have been gained by digital image processing. Image segmentation in image project which be taken as a crucial step transiting from image process in low level to image analysis in high level holds the important position in the research of image techniques. In the past, most efforts has been exerted in monochromatic image segmentation, but now, as the rapid improvement of processing ability of personal computer, the color image segmentation has attracts more and more attentions. However, most of the old segmenting methods applied to monochromatic image cannot be applied to color image segmenting directly, so, researches in developing new algorithms is necessary and the generality and better segmenting effects is the goal of researches.The paper has studied on color image segmentation technique, and the primary tasks of the paper are as follows:(1) The current and mainstream color image segmentation algorithms have discussed in two aspects. Firstly, the color spaces which have recently appeared in color image segmentation field have been analysed and summarized, then we concluded the characteristics of them when used in segmentation. Secondly, the mainstream segmenting algorithms have been analyzed, classified and summarized. Then we compared the advantages and disadvantages of them. At last, we point out that the effective way to segment the color image is taking the features of the specific segmenting problem into account, choosing appropriate color space and a suitable segmentation method corresponding to this color space.(2) Principal component analysis (PCA) is used in data compressing and statistical features extracting. And then, the basic theory of PCA is discussed and the corresponding experiments are done. Based on the analysis and experiments, we concluded that the fact that the direction of the vector related to the most value is the direction of the pixels'distribution can be applied to image segmenting.(3) On the basis of PCA and Fuzzy C-Means Clustering (FCM), a new algorithm which can be applied to segment the color image with shadows and faculae is proposed. We explained why we choose the fuzzy cluster method first. Then the limitations caused by space which data clusters in are analyzed, and'the desired image color eigenspace'used in the new methods are described in detail. The reason that we use...