The Research Of Distributed Compressed Video Sensing Based On Evolutionary Pursuit Algorithm And HMT

Due to the energy intake of video equipment battery and limited memory, they require low complexity of encoding algorithm;and for the decoder side, with large energy at the center of the computer, they can allow the high complexity of decoding algorithm.The demand and the traditional applications of video coding algorithm is just the opposite. In recent years, a grand new framewok of video coding-distributed video coding,DVC increasingly draws many attentions of researchers.With the aid of compressed perception theory framework of data acquisition, distributed video encoder complexity and power consumption will have reducing in the further, which makes it more effective. Current coding scheme of distributed compressive video sensing(DCVS) either require a complex feedback channel for effective operation or take advantage of the imformation inadequately. In this paper, a distributed Compressed Video Sensing codec based on hidden markov tree model(HMT-DCVS)is proposed. The hidden markov tree(HMT) model which will be able to exploit the distribution features of the wavelet coefficients of image is introduced into the residual frame reconstruction.To make full use of the structure characteristics of image wavelet coefficients, establish general hidden markov tree model of residual frame with side information(SI), and propose iterative termination criterion which contributes to decode adaptively that avoids using complex feedback information.Simulation results demonstrate that compared with the existing three kinds of methods, the proposed algorithm can reconstruct the video sequences faster and yet effective.Using evolutionay computation can guarantee the important characteristics of refactoring advantage, A new fast pursuit algorithm based on Bayesian framework and evolutionay computation is proposed in this thesis. In the proposed algorithm, a signal to be recovered is viewed as the sum of some signal components, the approximate reconstruction is achieved by iteratively estimating each signal component one by one. In Bayesian framework, the generalized Gaussian distributions with the different parameters are employed as the prior of signal components. The high-dimensional nonconvex optimization problem for CS reconstruction is decomposed into a series of low-dimensinal nonconvex sub-optimization problems. The evolutionary algorithm is adopted to solve the sub-problems to estimate the signals. The thesis designs a distributed video compression under the cognitive architecture of residual frame decoding method based on BEPA algorithm, and using Matlab software to the algorithm are simulated.