Research On Algorithm Of Color Image Segmentation Integrating Edge And Region Growing

The main task of image segmentation is to extract the interesting targets of the image, and divide the image into foreground and background. Color image segmentation is occupying an important position in the color image processing, since it is a precondition of pattern recognition and color image analysis. Many scholars have achieved good results on the study of color image segmentation method, but there is no segmentation method can be applied to the segmentation of all color images, so the research on color image segmentation and further improvement is significant.Edge-based and region-based segmentation methods are two kinds of common segmentation method. They get continuous development in the field of color image segmentation. This thesis aims at improving the segmentation accuracy, and proposes a method of color image segmentation integrating edge and region growing. In this thesis, the main work and innovations are as follows:1. Research on multi-scale module maximum edge detection method. Advantage of the characteristics of large-scale and small-scale edge detection, a multi-scale edge detection method is improved to achieve more complete edge with the research on single scale module maximum.2. Research on adaptive filter, which improves the segmentation accuracy by preserving image edge. The filtering method defines an adaptive filtering window which takes the relationship between neighborhood pixels into consideration. Experimental results show that the method can not only filter out the noise and smooth image, but also effectively protect the edge information.3. Research on an unsupervised color image segmentation method based on dual-tree complex wavelet transform. Extract accurate edge mapping by combining translational invariance and directional selectivity of DT-CWT with multi-scale properties of wavelet transform. Then adopt the dual mapping constraint criteria to generate initial seed points automatically, in order to increase the effective separator between large areas to make the areas’ edge more obvious. Finally, the implementation of region growing and iterative merging take into effects. This method greatly improves the segmentation accuracy, and reduces the cost of the man-machine combination, then the experimental results and the accuracy of the evaluation is further evidence of the effectiveness of the algorithm.