基于纹理信息的图像自适应分块压缩感知算法  

Adaptive block compressed sensing algorithm of image based on texture information

作　　者：左胤杰 赵君喜[1] ZUO Yinjie;ZHAO Junxi(School of Communication and Information Engineering,Nanjing University of Posts and Telecommunications,Nanjing 210003,China)

机构地区：[1]南京邮电大学通信与信息工程学院,南京210003

出　　处：《智能计算机与应用》2024年第7期128-135,共8页Intelligent Computer and Applications

摘　　要：针对传统分块压缩感知算法中对各图像块分配相同采样率,限制图像重构质量的问题,本文提出一种基于纹理信息的采样率自适应分块压缩感知算法。在观测端将图像一维灰度熵和标准差作为纹理信息量的衡量标准,使用K-means++算法将图像块按照纹理信息的相似性分为3类;结合边缘信息为3类图像块分配自适应采样率,并进行采样率二次分配;在重构端为缓解重构图像产生的块效应,采用分层分块结构结合改进的平滑投影Landweber算法进行重构。实验结果表明:在不同的采样率下,本算法重构图像在客观重构质量和主观视觉效果上均有一定提升,重构图像的块效应也得到缓解。To address the problem of assigning the same sampling rate to each image block in the traditional block compressed sensing algorithm,which limits the image reconstruction quality.A sampling rate adaptive block compressed sensing algorithm based on texture information is proposed.At the observation end,the method uses the one-dimensional grayscale entropy and standard deviation of the image as the measurement criteria for the amount of texture information,uses the K-means++algorithm to divide the image blocks into three categories based on the similarity of the amount of texture information,and then combines edge information to allocate adaptive sampling rates for the three types of image blocks,and performs a secondary allocation of sampling rates;At the reconstruction end,in order to alleviate the blocking effect generated by the reconstructed image,a layered block structure combined with an improved smoothed projected Landweber algorithm is used for reconstruction.The experimental results show that under different sampling rates,the reconstructed images using this algorithm have certain improvement in both objective reconstruction quality and subjective visual effect,and the block effect has also been alleviated.

关 键 词：分块压缩感知 自适应采样率 纹理信息 平滑投影Landweber算法 块效应 

分 类 号：TN919.8[电子电信—通信与信息系统]