Research On The Quantization Algorithm For Compressed Video Sensing

Compressed Video Sensing(CVS) is a novel video signal acquisition and encoding method. Compared with the traditional video encoding method with complex encoder and simple decoder such as H.26 X, CVS completes signal acquisition and compression simultaneously by random measure at encoding side, reconstructs the video frame by solving nonlinear problems at decoding side, transfers the computational complexity form the encoding side to the decoding side, has the great advantage of simple coding, suitable for computing power limited distributed scene such as wireless video surveillance etc..After deeply studied the CS theory and implementation framework, both uniform scalar quantization(SQ) and nonuniform scalar quantization, and an excellent compressed image sensing quantization algorithm,progressive quantization are researched in this paper, their advantages and disadvantages are analyzed through a large number of simulation experiments. Then, the CVS progressive quantization algorithm based on uniform quantization and nonuniform quantization are proposed. The main work of this paper include:1. Based on the correlation between frames of the video signal, this paper proposes progressive quantization method for CVS measurements. The measurements of non-key frames are quantized and only some less important bitplanes will be transmitted. At the decoder, neighbor reconstructed frames are used for motion estimation to generate side information of the non-key frames, then the estimation of the non-key frame measurements is obtained by measuring the side information frame. Finally, combined with the less important bitplanes transmitted from the encoder, accurate measurements are gotten by inverse quantization. Experimental results show that, compared to uniform scalar quantization, the proposed algorithm has a higher rate-distortion performance, without additional complexity in the encoder.2. Considering that the quantization error of nonuniform quantization is smaller than that of uniform quantization at the same quantization depth, this paper proposes an adaptive nonuniform quantization algorithm based on probability density function, then combining with progressive quantization, proposes CVS nonuniform progressive quantization method. Experimental results show that the propoesd method has a higher rate-distortion performance compared with CVS progressive quantization algorithm.