摘要
探讨了神经网络用于图像压缩和解压缩技术,实现了一种基于分类的改进BP神经网络压缩方法:按图像各部分像素特征将图像分为平滑块、中间块和边缘块,不同的分类块采用不同的隐含层数,从而在保证重建图像丰富细节的同时,提高图像的压缩比.同时,对3层BP神经网络进行优化,提高了网络的收敛速率,实验结果证明本算法在取得较大的压缩比的同时能保证图像的恢复质量.
Neural network used in image compression and decompression is discussed in this article. Then an image compression method based on improved BP net.work is developed. In this article we classify an image into three seetions according to its pixel character. The BP algorithm is also bettered to improve the convergence velcity of BP neural network and the simulation results indicates that the compression rate and the decompression quantity is largely improved.
出处
《兰州大学学报（自然科学版）》
CAS
CSCD
北大核心
2005年第4期70-72,共3页
Journal of Lanzhou University(Natural Sciences)
基金
985特色项目计划支持课题(LZ985-231-582627)甘肃省自然科学基金(ZS001-A25-008-Z)资助项目.
