基于神经网络的压缩感知图像重构算法

Compressed sensing image reconstruction algorithm based on neural network

压缩感知理论相较于传统采样定律具有压缩率小的巨大优势,探究更优的信息重构算法具有重要意义。为解决传统压缩图像恢复时间长、压缩率高的问题,提出以循环神经网络进行图像恢复的方法,并将其与全连接网络、卷积神经网络比较分析,实现了2.2%压缩率下128×128单通道图片的重构。研究中,进一步比较分析了3个数据集、3种输入维度下,网络算法的性能指标及单张图像的恢复效率,证明了神经网络算法对图像重构的有效性、实时处理视频的可行性。

Compressed sensing theory has a huge advantage of small compression ratio compared with the traditional sampling law,and it is of great significance to explore better information reconstruction algorithms.In order to solve the problem of long recovery time and high compression rate of traditional compressed images,the method of image restoration using recurrent neural network was proposed and compared with fully-connected network and convolutional neural network to achieve the reconstruction of single-channel images of 128×128 with 2.2%compression ratio.In the research,the performance indexes of the network algorithm and the recovery efficiency of the single image were analyzed under three data sets and three input dimensions.The validity of the neural network algorithm for image reconstruction and the feasibility of real-time processing video are proved.