Image Compression Based on the Improved SA Algorithm
Aiming at the techniques of the image compression, we put forward a kind of improved local learning adaptive Silva-Almeida (SA) algorithm on the basis of BP neural networks. The training process of the new algorithm was divided into the acceleration and steady convergence periods,and we adjusted the learning rate and the momentum factor locally at the same time. By the experiment of the curve fitting, the improved SA algorithm quickened the training speed, raised precision distinctly, and resisted the oscillation in a large scale. In the test of the image compression,compared with the resilient propagation (RPROP) method,the new algorithms training speed and the quality of the reconstruction images were enhanced very greatly, and we acquired a higher ratio of image compression. Therefore the improved SA algorithm is more effective to the image compression.
SA algorithm image compression BP networks resilient propagation neural networks
Yiqian Tang Chunli Feng Yue Zhao Lin Shi
Chengdu University,Chengdu, 610106, China Liaoning Institute of Technology, Jinzhou, 121001, China
国际会议
杭州
英文
361-365
2006-10-12（万方平台首次上网日期，不代表论文的发表时间）