Compressive sensing image reconstruction based on stage-wise directional pursuit searching in the WBCT domain

In the past, it is the Nyquist sampling theory the central principal of signal processing, and most of the data we acquire can be dropped off with nearly no perceptual loss. However, an emerging theory named "compressive sensing" has been taking the place, which shows that super-resolved images can be reconstructed from extremely fewer measurements than what is generally considered necessary. In this paper, under the theoretical compressive sensing framework, we recommend the optimally- jittered undersampling, the wavelet-based contourlet transform and the stage-wise directional pursuit as the three main ingredients -a sampling strategy, a sparse transform and a recovery algorithm to recover images with and without noises. Some standard gray and color images are tested in the experiments, turning out that we only use measurements that around 40 percent of the original size to reconstruct images within high PSNR value and favorable vision impression, which means that we will dramatically reduce measurement time, sampling rates, and the use of analog-to-digital converter resources in certain corresponding practice domain.