Soft threshold iteration-based anti-noise compressed sensing image reconstruction network

Optical images of artificial satellites can provide wide-range geographic information, but their large amount of information and severe noise interference during transmission limit their applications in strategic deployment land, resource census and other fields. In this letter, the Soft Threshold Iteration-based Anti-noise Compressed Sensing Image Reconstruction Network is proposed to address the problem. The network proposes a reconstruction denoising hybrid network, employs adaptive factors and Gaussian initialized unconstrained adaptive sampling matrix, and proposes a reconstruction network mean square constraint between stages. According to experiments, the network can achieve a maximum peak signal-to-noise ratio of 37.43 dB when the sampling rate is 50% and the measurement values are mixed with Gaussian noise with a mean of 0 and SNR 30 dB.