Distributed Video Compressed Sensing Secondary Reconstruction Based on Inter-Frames Similarity Structure

Abstract

The emergence of distributed video compressed sensing technology has effectively reduced the computational cost and power consumption of the video system. As the core content of the DCVS research, reconstruction algorithms are mostly focused on improving the reconstruction quality of CS frames, while neglecting the reconstruction quality of the key frame. The key frame is not fully utilized as well. Focus on these problems, the paper proposes a distributed video compressed sensing secondary reconstruction algorithm based on similar structure between frames. In order to eliminate the influence of non-overlapping block CS that may introduce noise and block effects on key frames, firstly, the fractional total variation algorithm is used to initially reconstruct the key frames, which improves the reconstruction quality while reducing the computational complexity. Then, a CS frame secondary reconstruction algorithm based on structural similarity is proposed, which uses SSIM as the matching criterion to search for matching block groups similar to the current frame from multiple reference frames, as a regular term to achieve the secondary reconstruction of CS frames. The experimental results show that the overall performance of the proposed algorithm is better than the current advanced DCVS method. The research results of the thesis can be used for reliable compressed sensing reconstruction of video and image.