燕山大学学报2024，Vol.48Issue(2)：157-164,8.DOI:10.3969/j.issn.1007-791X.2024.02.006
基于聚类与稀疏字典学习的近似消息传递
Clustering and sparse dictionary learning based approximate message passing
摘要
Abstract
Dictionary learning based approximate message passing(AMP)has a high demand on the number of training samples,and its computational cost is high.The double sparse model is introduced to study sparse dictionary learning based AMP,which reduces the demand on the number of training samples in the iterations and improves imaging quality and efficiency.Furthermore,the clustering and sparse dictionary learning based AMP is proposed.In iterations,the clustered blocks are denoised adaptively with sparse dictionary learning.In comparison to traditional dictionary learning based AMP,the clustering and sparse dictionary learning based AMP can achieve 0.20～1.75 dB higher peak signal-to-noise ratio of the reconstructed images,and improve the computational efficiency by 89%in average.关键词
图像重构/近似消息传递/字典学习/稀疏字典/聚类Key words
image reconstruction/approximate message passing/dictionary learning/sparse dictionary/clustering分类
信息技术与安全科学引用本文复制引用
司菁菁,王亚茹,王爱婷,程银波..基于聚类与稀疏字典学习的近似消息传递[J].燕山大学学报,2024,48(2):157-164,8.基金项目
河北省自然科学基金资助项目(F2021203027) （F2021203027）
燕山大学基础创新科研培育项目(2021LGZD011) （2021LGZD011）
河北省重点实验室项目(202250701010046) （202250701010046）
