A Reconstruction Method for Video Compression Sensing

Abstract

The invention discloses a reconstruction method for video compressed sensing. Images of frames are reconstructed independently after a receiving end receives measurement values of the frames; motion estimation and motion compensation based on blocks are carried out among the reconstructed images to obtain predicted images of the frames; reconstruction problems are built for the images of the frames by means of the predicted images and the measurement values, and the reconstruction problems are divided into five sub-problems, namely a single-frame image reconstruction problem, a residual image reconstruction problem, an image gradient variable reconstruction problem, a residual gradient variable reconstruction problem and an image orthogonal basis coefficient variable reconstruction problem, for alternant iteration solving. Every time the five sub-problems are solved through alternant iteration solving, image reconstruction quality is further improved by the adoption of a bivariate shrinkage threshold method. If inner loop end conditions are met, the iteration solving process of each frame of image is finished; if the upper limit of outside loop iteration times is reached, the construction procedures are over. By means of the method, performance and the effect of image reconstruction can be effectively improved.

Description

technical field [0001] The invention relates to compressed sensing technology, and particularly designs a video compression sensing reconstruction method in the compressed sensing technology. Background technique [0002] Compressed Sensing (CS) is an emerging theoretical system that has received extensive attention in the field of signal processing in recent years. The theory states that if a signal to be sampled is sparse (or sparse in a transform domain), then the possibility of accurately recovering the original signal from its limited uncorrelated measurements is extremely high. The introduction of CS theory has greatly enriched and developed the signal acquisition theory, and provided new technologies and new ideas for research in other related fields. Because the computational complexity of the measurement process of CS is low, and the reconstruction process needs to iteratively solve the optimization problem with relatively high complexity, its theoretical structure...

AI Technical Summary

Problems solved by technology

Since the residual reconstruction method only considers the temporal correlation and ignores the spatial structure characteristics of the image itself, if the obtained predicted value is inaccurate, the quality of the reconstructed image will be reduced

[0008] In summary, at the receiving end, in order to improve the quality of independently reconstructed images, it is necessary to make full use of the correlation between adjacent frames; the performance of existing methods needs to be further improved

Images