The Research Of Video Image Reconstruction Algorithm Based On Adaptive Block Compressed Sensing

Compressed Sensing(CS)theory breaks through the limitation of traditional video image signal sampling theorem,its sampling rate is much lower than the Shannon-Nyquist sampling rate and satisfy sampling and compression at the same time.CS is applied to signal processing to avoid collecting large amounts of redundant data,thereby effectively saving the system time and resources.On the basis of CS theory,L Gan et al.Proposed the theory of Block Compressed Sensing(BCS),the main idea of BCS is to divide the original image into several image blocks of equal size,and then independently measure each image block with a fixed number of samples.Finally,each image block is reconstructed and spliced into the original image.After the BCS is presented,the large memory and high computational complexity of the video image in the reconstruction process are largely avoided.In the process of using BCS theory to deal with video images,the reconstruction algorithm is the core link,which is directly related to the final reconstructed video image quality.But now,most of the reconstruction algorithms based on BCS framework have a few shortcomings such as insufficient image block sampling and high computational complexity.Therefore,it is necessary to study the effective adaptive sampling method and the high efficient reconstruction algorithm.Based on the above analysis,the main work of this paper is as follows:(1)On the basis of BCS,we analyzed the characteristics of wavelet coefficients and variances,which represent the details of the image.And an adaptive sampling scheme based on wavelet coefficients and variances were developed.On the basis of this,two kinds of image block adaptive CS methods under different sampling modes were proposed,which were image block adaptive CS based on wavelet coefficients and image block adaptive CS based on variance.(2)Based on the statistical information of adaptive BCS,for the shortcomings of Iterative Shrinkage Thresholding(IST)algorithm in convergence rate and reconstruction precision,an improved BCS-IST algorithm was proposed by replacing the1 l norm with the Total Variation(TV)as a regularization constraint.The experimental results showed that the proposed algorithm can better preserve the structure and texture information of the image to obtain better image reconstruction effect as compared with the traditional BCS-IST algorithm.Moreover,the algorithm accelerates the convergence rate of the algorithm to a certain extent.(3)Through further analysis of the temporal and spatial correlation of video signals,an adaptive BCS video reconstruction method based on space-time feature was proposed.Firstly,the measurement matrix suitable for the video signal was constructed by the Kronecker product operation,which achieved the overall compression measurement of the video signal.Then,the Multiple Hypothesis(MH)model was combined with the minimum TV model to construct the predictive-residual reconstruction model of joint space-time feature,and the prediction frame of the current frame was obtained by iteration.Finally,the improved BCS-IST algorithm was used to calculate the residuals of each block and combined with the prediction frame of the current frame to reconstruct the current frame.The experimental results showed that the proposed method can further reduce the computational complexity and speed up the algorithm operation speed while improving the quality of video reconstruction.