Object segmentation based on multi-resolution texture analysis

This paper proposes the object segmentation method using multi-resolution texture analysis. The method consists of 1) pre-processing to produce multi-resolution images, 2) texture analysis using 1-Nearet Neighbor and Neural Networks, and 3) post-processing to combine the segmentation results. This structure is proposed in order to cope with the variety of the textures. The experiments using real test images prove that this multi-resolution approach could solve the cases with variety of the textures efficiently, and show that the method could achieve about 77% segmentation accuracy on average. The future study includes the application of the Liner Regression and an examination of some feature based approach