Color Image Fusion Segmentation Methods Based On Multi-feature

Image segmentation is an important technology in image processing, and has been widely applied in traffic, medicine, agriculture, industry, and so on. It is also a classical problem in computer vision which has not been well solved. There are many methods of image segmentation, however, most of them only can be used in specific images, do not have universal applicability and validity. In recent years, some researchers introduce information fusion strategy to image segmentation in order to improve the segmentation effect. The research of color image fusion Segmentation method has broad application prospects.This thesis includes the following contents:(1) We introduced the definition and the general process of image segmentation, summarized the main method of image segmentation.(2) We introduced the definition and basic principle and hierarchical structure of information fusion, summarized the method of image segmentation based on fusion, described the segmentation method based on feature fusion and the segmentation method based on multi-scale fusion in detail, collated the classical evaluation criteria of image segmentation.(3) In order to solve the problem of presenting complex scene information in a specific single color space, we use a method based on hierarchical clustering to fuse multiple segmentation results of multiple color space. We carried out segmentation experiments on Berkeley segmentation database and compared with a variety of classical segmentation methods. The experimental results indicated that this segmentation method can get higher segmentation accuracy, and had advantages for overcoming over segmentation problem.(4) According to the segmentation evaluation criteria PRI, we derive a fusion model for combining multiple segmentation results, through minimizing the Gibbs energy function of this model to obtaine the optimal fusion result. We carried out a variety of experiments on Berkeley segmentation database and compared with other classical segmentation methods. The experiments indicated that the segmentation results of this method were more consistent with the ground truth.