An adaptive-anisotropic-diffusion filtering is a method used to filter images, with an adaptive parameter K for coarseness. In the filter process, the parameter adjusts the diffusion process to make a highly-textured region smoothed. In addition, the image noises can be removed, and the edge can be kept sharp. A region growing method is also used to segment the image. Normalized-combined-distance method (color and coarseness distances) is introduced to measure the similarity between two adjacent regions. This method is convenient for selecting region with merging threshold. Finally some segmentation results are compared with the ones by JSEG method. The experiments show that such method is a good approach for image segmentation.
展开▼