Multi-scale Image Segmentation Algorithm Based on SPCNN and Contourlet
A new multi-scale image segmentation algorithm based on nonsubsampled contourlet transform (NSCT) and simplified plus coupled neural network (SPCNN) has been discussed in this paper. Comparing with plus coupled neural network (PCNN), the SPCNN algorithm can decrease the complexity of adjusting parameters significantly. First we combine susan edge detector with SPCNN, more accurate result can be obtained. Then we use SPCNN to deal with the low-frequency coefficients of NSCT, and then the running time will be shortened remarkably. In order to solve the problem that the details of the image will be fuzzed because of losing high-frequency coefficients of NSCT, we preserve the edge information in corresponding high-frequency coefficients by detecting the edge of origin image. Finally, we use maximum mutual information (MMI) to determine optimal results by SPCNN. The test results prove the rationality of this method and show efficiency and accuracy to a certain extent.
Nonsubsampled contourlet transform pcnn susan edge detector mazimum mutual information
Dongfang Chen Tao Xu
College of Computer Science & Technology Wuhan University of Science & Technology Wuhan,China
国际会议
上海
英文
2964-2968
2009-11-20（万方平台首次上网日期，不代表论文的发表时间）