Research On Coding Scheme For 3D Holoscopic Image

3D holoscopic imaging can provide continuous motion parallax throughout the viewing zone with precise convergence and depth perception as well as supply a more real and natural 3D visualization. It can also minimize some uncomfortable feelings such as eye strain or headache when people focus on the screen for a long time. For this reason, it is regarded as a promising technique for future 3D video technology. At the same time, due to the focus-after-shoot, extendable depth of field and switchable points of view abilities, the 3D holoscopic imaging is attracting significant interests. In order to represent the captured 3D holoscopic image with adequate resolution, large amount of data is required, which is more lager than the natural 2D image. Consequently, effective compression schemes become of paramount importance for such particular type of content. However, the coding efficiency of the most existing 3D holoscopic image coding method is not very high with a high computational complexity. Moreover, most of them can not provide a scalability of the 3D holoscopic image. In order to alleviate the shortcomings, in this dissertation,the following aspects have been done to further improve the coding efficiency of the 3D holoscopic image.Firstly, In order to alleviate the shortcoming that the existing encoders, such as JPEG, H.264 and HEVC, can not achieve a high coding efficiency, we proposed a disparity compensation based 3D holoscopic image coding algorithm with the high self-similarity of 3D holoscopic image. The proposed method tries to improve the coding efficiency by estimating the disparity between the neighboring element images. However the disparity compensation based coding algorithm can not achieve a better prediction on some texture and edge regions. To sove this problem, we fuether proposed a hybrid disparity compen-sation based prediction and intra block copy prediction algorithm. With the rate-distortion optimization being adopted to choose the optimal block partition and the prediction modes,the proposed hybrid algorithm can choose the prediction method adaptively. Experimental results show that the proposed method can improve the coding efficiency and reduce the computational complexity at the same time.Secondly, in order to further improve the prediction accuracy on some texture and edge regions and protect the texture details of the constructed element images, we proposed a Gaussian process regression based coding method and a kernel density estimation based coding method. The Gaussian process regression based coding method tries to construct the current coding block and its prediction support as a Gaussian process. Incorporating kernel functions, the prediction support is projected into a high-dimensional feature space to fit the anisotropic and nonlinear image statistics. Instead of directly conditioned on the support, Gaussian process regression is leveraged to make prediction in the feature space. The kernel density estimation based coding method tries to construct a statistical model and utilize kernel-based MMSE (K-MMSE) estimation to predict the coding block.Experimental results show that the Gaussian process regression based coding method and the kernel density estimation based coding method can outperform the HE VC intra standard with 2.67dB and 2.65 dB quality improvement, respectively. And they all can achieve a better visual quality than HEVC, especially in some texture regions.Thirdly, in order to improve the coding efficiency by exploring the spatial correlation among the view images with different perspectives rendered from 3D holoscopic image,we proposed a spatial correlation based 3D holoscopic image coding method. And in order to avoid the transmission delay, we further proposed a scalable coding method of 3D holoscopic image by using a sparse interlaced view image set and disparity map.In the spatial correlation based 3D holoscopic image coding method, we proposed to explore the high spatial correlation between the neighboring view images. And in the 3D holoscopic image scalable coding method, we try to use the sparse interlaced view image set and disparity map to characterize the origin interlaced view image approximately to provide the scalability of the 3D holoscopic image. Experimental results show that the proposed method can achieve a better coding efficiency and provide the scalability of the 3D holoscopic image at the same time.