An improved block matching and prediction algorithm for multi-view video with distributed video coder

• Main Authors: Ching-hua Chen, 陳慶樺
• Other Authors: Jiann-jone, Chen
• Format: Others
• Language: zh-TW
• Published: 2010
• Online Access: http://ndltd.ncl.edu.tw/handle/52392853740216273434

碩士 === 國立臺灣科技大學 === 電機工程系 === 98 === With the advance of video coding technology, the multimedia platform can provide Three-dimensional Television (3D TV) display for users with or without glasses. The inter-view video correlation has been exploited to perform efficient multi-view video coding. For the extension applications on mobile devices and wireless networks, the distributed video codec (DVC) can be utilized to shift encoder complexity to decoder under the multi-view video coding framework, denoted as Multi-view Distributed Video Coding (MDVC). Several coding control methods have been proposed to improve the MDVC codec performance: (1) The block transform coefficients are utilized to design the priority rate control for the turbo coder at decoder. (2) A Block Matching and Prediction algorithm, based on Scale Invariant Feature Transform, abbreviated as SIFT-BMP, has been proposed to yield more accurate side information for better DVC reconstructed images. In comparisons with other view estimation method, the proposed method can provide lower computational complexity and time consumption, while yielding better reconstructed video quality. Simulations show that the proposed SIFT-BMP method can reduce the time complexity of motion estimation to 1.29~2.56 times smaller, as compared to previous Hybrid Multi-View Motion Estimation (H-MVME) methods, while the PSNR of decoded video can be improved 0.9~3dB, as compared to H.264/AVC intra coding.