The Study Of Color Image Segmentation Method Based On Primary Characters Adaptive Selection

Image segmentation is a fundamental and important technology in image engineering. It is the primary task in any image analysis and image recognition process because all the subsequent jobs such as feature extraction, object recognition etc rely heavily on the quality of image segmentation. In recent years, with the development of technologies, such as machine vision, pattern recognition and content-based image retrieve etc and the wide usage of color image, image segmentation, especially color image segmentation, has played a more and more important role.In this paper, a color image segmentation algorithm is proposed which is based on traditional color image segmentation technology. It is the study of color image segmentation based on Primary Characters adaptive selection. The segmentation outcome is kept consistent with vision psychology perception. So color space with vision-consistent is chosen. HSL space was one of the widely used color spaces. HSL is an intuitionistic color space. Each component in HSL corresponds to hue, saturation and light of vision, which fits color image segmentation well. But the singularity in HSL space is an inevitably problem. To overcome the singularity problem and keep algorithm consistent, we introduce the concept of color purity. We use color purity as weight of hue in singular point. With the occurrence of strong or weak light etc affected by image illumination and visual angle difference, hue becomes meaningless. We determine Primary Characters component by the times of smoothing histogram to unimodal. This method can avoid errors caused by Primary Characters segmentation with experience and fixed hue. For most of natural images, their histograms are sparser which shows unobvious peak -valley character. With high noise or other factor influence, the peak-valley character of histogram in attribution space becomes fuzzy which cause some problems in segmentation. In this paper, we process the histograms of hue and light with fuzzy enforcement, which can heighten peak and valley character. For fuzzy enforcement histograms, we label the pixels by Primary and inferior Characters, which not only can reduce false segmentation error, but also can get good clustering for all pixels. We give the new region distance function, which combines the regions until conditions are met. The gained regions are final segmentation outcome.To validate the performance of algorithm, we compare the proposed algorithm with other algorithm with many images by means of two aspects: image segmentation effect ion image, measurement function of segmentation evaluation. Experiments show that the proposed method has good segmentation outcome.