摘要
针对图像缩放中的问题,提出基于自联想记忆算法。首先确定自联想记忆最小上界和最大下界,自联想记忆每个层之间的神经元通过最近域互连接方式连接信息交替,能够在神经系统缺损信息时通过自联想记忆恢复出原来存储的完整记忆模式;接着所有神经元节点采用同步方式调整状态,利用均场定理加快自联想记忆神经网络学习速度;最后给出了图像缩放算法过程。实验仿真显示算法输出图像能够较好地保持原图像内容,峰值信噪比(PSNR)比较大。
Image scaling proposed algorithm is based on self-associative memory.First determine associative memory least upper bound and greatest lower bound,since associative memory between each layer neurons interconnect recent domain connection information alternately nervous system defects by self-associative memory to restore the original storage complete memory mode;then all neurons nodes using synchronous adjustment of status,the use of the mean field theorem to accelerate from associative memory neural network learning speed;Finally,the process of image scaling algorithm.The real experimental simulation shows that the proposed algorithm can output images to keep the content of the original image,PSNR large.
出处
《科学技术与工程》
北大核心
2013年第18期5381-5384,共4页
Science Technology and Engineering
关键词
自联想记忆
神经网络
学习
缩放
associative memory neural networks learning scaling
