Study On The Reconstruction Algorithm Of Video Image Based On Compressed Sensing

The convergence speed of existing compressed sensing reconstruction algorithmapplied to the video image reconstruction is slow. In order to meet the real-timerequirements of video image reconstruction on the basis of the existing image processingalgorithms, some fast compressed sensing reconstruction algorithm will be researched inthe paper, the main work is as follows:The fast iterative shrinkage threshold algorithm is introduced into the Compressedsensing image reconstruction and there is some improvements based on compressedsensing structural features. The experimental results show that the reconstructed imagequality is good, and convergence speed of the reconstruction algorithm is improvedsignificantly.The compressed sensing reconstruction algorithm based on the split augmentedLagrangian shrinkage threshold is discussed. This algorithm takes advantage of theobjective function the second derivative while introducing the idea of variablesegmentation, a single spatial domain variables are divided into the spatial domain andfrequency domain variables, which constrains each other, and then approaches theoptimal solution alternately. The experimental results show that this algorithm has aobvious advantage in terms of both the quality of reconstruction image and convergencerate.In order to solve the problem that some image detail is lost beause the samethreshold for all wavelet coefficients shrinkage is used in the compressed sensingreconstruction algorithm ,a improved split augmented Lagrangian shrinkage thresholdalgorithm is receive to complete the compressed sensing reconstruction. This improvedalgorithm overcome the shortcomings that soft threshold shrinkage is over smooth andretain more image detail. The experimental results also proved the effectiveness of thealgorithm.A video reconstruction of compressed sensing algorithm based on frame differenceis discussed. The background image of adjacent frames which is a priori knowledge of the next frame image reconstruction is gained by the frame difference method based onthis characteristic. The video image is reconstructed by frame difference method andreconstruction algorithm mentioned in the paper. The time-consuming of achieving thebest reconstruction image is reduced and the quality of the reconstructed image isimproved to some extent. Experimental results show that compared the frame differencecompressed sensing reconsturction algriothm with the original image reconstructionalgorithm, the quality of the reconstructed image is improved both in subjective visionand objective peak value signal to noise, while the convergence rate of the algorithm isalso increased significantly.