Color Image Segmentation Algorithm

Image segmentation serves as the key of image analysis and pattern recognition. It's a process of dividing an image into different region such that each region is homogeneous, but the union of any two regions is not. Early researches mainly focus on gray-scale images with more sophisticated algorithms. With the constantly updated computer equipment, as well as the advancement of technology, there are more and more color images of various applications, and color image segmentation is also causing more and more attention. However, color image segmentation is more complex than gray-scale images. In this dissertation, a color image segmentation algorithm based on histogram, FCM clustering, as well as region combination (HRFCM) is proposed. Experiments prove that HRFCM performs well on many color image segmentations.The main job of this dissertation is as follows:1.The selection of color space in color image segmentation is researched.2.The image is divided into non-singular points and singular points, which is transformed from RGB space to HSV space. Non-singular points and singular points are separately clustered. This process not only makes the partition consistent with the human visual psychology, but also overcome the singularity of HSV space.3.The FCM clustering algorithm is the combination of fuzzy math and cluster analysis, so that algorithm in dealing with the problem generated by fuzzy uncertainty in image segmentation is more in line with the actual situation.4.Characteristics of the image pixel is mapped to the one-dimensional histogram , thus significantly reduce computational complexity and greatly improve the speed of the algorithm.5.The histogram peaks selection algorithm is introduced to determine the number of the clusters and the initial cluster centers, thus get more accurate cluster numbers and cluster centers to reduce the iteration number and cluster into local extremum possible, while automatically dividing images without manual intervention.6.The use of image spatial information for regional mergers eliminates the scattered small area after clustering, which overcomes the over-segmentation problem in FCM, and increases the ability of anti-noise.7.The validity of HRFCM algorithm is tested through a lot of experiments on color image segmentation, and the performance between HRFCM and other common algorithms has been compared.