Study On Color Image Segmentation Tree Matting Algorithm

Color tree image segmentation is tree image visualization, tree growth condition assessment of the theoretical basis, due to the complexity of color tree image and background, the tree image segmentation is relatively difficult, this paper on color tree image segmentation method research, the main contents are as follows:Based on the traditional two-dimensional maximum entropy, watershed based color tree image segmentation method, in the simple background and complex background color tree image segmentation research, and segmentation results analysis.The traditional Bayes matting algorithm theory research, according to the cover as the value of the solution is not accurate, mask like graph edge is not smooth, put forward improved Bayes matting algorithm, join the smoothness constraint, obtains the final mask value is closer to the actual value and on two experimental results analysis and comparison.Based on the traditional Poisson matting algorithm theory research, the image into three for regional growth, reduce the unknown pixels in the region, to achieve the improvement of traditional Poisson matting algorithm, reduce the image segmentation arithmetic time, two groups of experimental results are compared.On robust matting algorithm for theoretical study and experimental. In view of traditional algorithm mask like value solely by sampling foreground and background samples to determine, leading to mask like value discontinuous problems, mask like the value of local continuity, the function optimization based on to achieve improvement of traditional robust matting algorithm, and two groups of segmentation results were compared.The tree image in the image database using the above three kinds of improved algorithm for the segmentation of the comparison and analysis of experimental results,comparison algorithm segmentation process complexity, the integrity of the time and the segmentation results to verify the feasibility of the improved algorithm.