MC-ISTA-Net: Adaptive Measurement and Initialization and Channel Attention Optimization inspired Neural Network for Compressive Sensing

TL;DR

This paper introduces MC-ISTA-Net, a neural network for compressive sensing that adaptively optimizes measurement, initialization, and channel attention to improve natural image reconstruction.

Contribution

It proposes an adaptive measurement, initialization, and channel attention mechanism within an optimization-inspired neural network for enhanced CS reconstruction.

Findings

Improved reconstruction accuracy over existing methods

Effective adaptive measurement and initialization strategies

Enhanced feature channel attention for better image recovery

Abstract

The optimization inspired network can bridge convex optimization and neural networks in Compressive Sensing (CS) reconstruction of natural image, like ISTA-Net+, which mapping optimization algorithm: iterative shrinkage-thresholding algorithm (ISTA) into network. However, measurement matrix and input initialization are still hand-crafted, and multi-channel feature map contain information at different frequencies, which is treated equally across channels, hindering the ability of CS reconstruction in optimization-inspired networks. In order to solve the above problems, we proposed MC-ISTA-Net