Research Of Image Segmentation Based On Color Space

Image segmentation is not only the basis of image processing technology, and is also oneof the most difficult research projects in image processing vision research. The quality of theimage segmentation determines the effect of target detection, feature extraction and targetrecognition in the image understanding process. It means that the quality of imagesegmentation determines the final image quality analysis and pattern recognition results;therefore, image segmentation has extremely important significance in image processing.Fuzzy clustering image segmentation is a commonly used method in image segmentation,which combines the advantages of the fuzzy set theory and the cluster theory. Fuzzy C-meansclustering algorithm (FCM) is the most representative method. As a kind of unsupervised fuzzyclustering method, it can avoid the human intervention in the implementation process andmakes the algorithm has better stability in the image processing. Fuzzy C-means clusteringalgorithm has been successfully applied in the field of image segmentation, data mining, targetrecognition, medical diagnosis, etc.Fuzzy C-means clustering algorithm is the main research object in this paper. The authorfirstly introduces the hard c-means clustering and the traditional fuzzy C-means clusteringalgorithm. In order to overcome the shortcoming of the traditional fuzzy C-means clusteringalgorithm, we make improvements in two aspects:(1) we firstly use the hierarchical subtractionclustering to get fast fuzzy C-means clustering, in order to get based on HIS color space fastfuzzy c-means clustering algorithm, we combine the fast fuzzy C-means clustering methodwith HIS color space. Experiment results show that the improved algorithm can increase thecalculating speed.(2) We calculate the high weight set of the image pixel and choose theappropriate initial cluster center according to the weight, make fuzzy cluster analysis to get thesegmentation result, then put forward the based on color space and optimization of the center ofthe initial fuzzy C-means clustering algorithm and related experiments and results analysisindicate that the new proposed method is effective. Finally, the author summarizes the mainworks of this paper and point out the next step of work and focal point of research.