xingwangsfu / mr-amp

Multi-resolution approximate message passing algorithm for multi-resolution compressed sensing problem

MR-AMP

MR-AMP algorithms

Introduction

This toolbox is to implement the multi-resolution compressed sensing reconstruction via approximate message passing (MR-AMP) proposed in the paper "Multi-resolution compressed sensing reconstruction via approximate message passing" (http://arxiv.org/abs/1508.02454).

Usage

Two different multi-resolution approaches are proposed, i.e., transform domain and spatial domain.

• Transform domain: For transform domain approach, please find the demo file "DCT_WT_Demo.m" in the "demo" folder.

• Spatial domain: For spatial domain approach, please find the file "TVAMP_Demo.m" in the "demo" folder.

• Comparison with TVAL3: For the comparison between TVAL3 and AMP-TV-2D, please refer to "tval3_vs_tvamp.m" in the "demo" folder. You can download TVAL3 toolbox here http://www.caam.rice.edu/~optimization/L1/TVAL3/ , and need to replace the TVAL3.m in the “solver” folder in the TVAL3 toolbox with TVAL3.m in our "solver" folder, since we aim to find the optimal slack parameters for TVAL3.

mr-amp's Issues

some problem with your codes

Hi Xing, Very thanks for your work. I read your paper and implemented your code. and the result became very similar to the result of your paper. However, I have some problems with your code which are as follows:
1- I couldn't find any code for Noise sensitivity of MR-AMP-ST (Table I). do you have its code or would you please describe further that how you got the values of noise sensitivity for empirical MR-AMP? I did not understand that where the value of γ is used? and why the noise sensitivity for empirical LR-AMP Is dependant to γ?
2- when I run tval3_vs_tvamp.m it errors! However, I didn't change your code!