抄録

H-044
Object segmentation based on multi-resolution texture analysis

This paper proposes the object segmentation method which consists 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 could cope with the variety of the texture in some extent. The experiments show that the method achieved about 77% segmentation accuracy on average. The future study includes the application of the Liner Regression and an examination of some feature based approach.