Depth Map And 3D Video Coding Based On Compressive Sensing

Recently, with the widely application of the three-dimensional (3-D) video, the functionality with user interaction and unique stereoscopic impression of 3-D video have drawn significant attention among industry and academic researchers, and the 3-D video has been widely used in education, entertainment, medical treatment, business, remote video monitoring and many other areas. It will play an important role in the future of video communication field. However, with the increasing number of the viewpoint, the volume of 3-D video data is growing. In order to avoid the transmission problem and the surge of the data in channel, the data must be compressed effectively. The theory of compressive sensing (CS) successfully broke through the limitation of the traditional temporal sampling theory that the signals are sampled at a rate at least twice the highest frequency in the signals. CS indicates that the sampling frequency no longer depends on the bandwidth, and will depend on the structure and content of the signal. It processes a new way to solve the problems above. In this paper, CS is used in the coding of depth map and 3-D video, and the main works are as follows:(1) Since the depth map meets the condition that the signal is sparse, we propose a novel depth map coding scheme based on adaptive block compressive sensing. When the image is compressed and reconstructed based on CS theory at the same sampling rate, so the low sampling rate can’t ensure that every block in the image obtains the well reconstruction quality and the high sampling rate usually leads to the resource waste due to the different sparseness degree of each block in the same image. In order to solve the problem above, an adaptive depth map coding algorithm based on CS is proposed in this paper. In the scheme, the proportion of the edge for each image block is used as the principle to estimate the sparseness degree, then according to the sparseness degree, different sampling rates were chosen adaptively for different sizes of block, thus a higher reconstruction quality is acquired for the depth map at a lower sampling rate;(2) The traditional compressive sensing uses a square block mode, but we found that there are some unreasonable places in the mode at the edge of the block. So we expect a more detailed block method which divides the whole depth map to eight block modes depending on the entropy of the image block. In this paper, based on the entropy of each block, we propose a depth map coding scheme based on adaptive rectangular block compressive sensing to obtain the higher reconstructed image quality.(3) The conventional video coding method, due to its high coding complexity, is not conducive in the resource-constrained mobile terminal applications. Distributed video coding and compressive sensing are used in the system of multi-view video coding. Video frames in each view are split into key frames and CS frames. Then they are encoded independently and decoded jointly. Thus we can receive the reconstructed image with high quality and low coding complexity.