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

基于传统字典学习的近似消息传递(approximate message passing,AMP)算法对训练样本数量的需求较高,且运算成本较高。本文引入双稀疏模型,构建基于稀疏字典学习的AMP框架,降低迭代过程中字典学习对训练样本数量的需求,提高压缩感知图像重建的质量与效率。进一步,提出基于聚类与稀疏字典学习的AMP算法,在迭代过程中依据图像块特征进行分类,并为各类图像块分别学习稀疏字典,实现自适应去噪。与基于传统字典学习的AMP算法相比,基于聚类与稀疏字典学习的AMP算法能够将重建图像的峰值信噪比提高0.20~1.75 dB,并且能够将运算效率平均提高89%。

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.