A Novel Data-Embedding H.264/AVC Coding Scheme for Error Resilience

• Main Authors: Kun-Chin Han, 韓昆瑾
• Other Authors: Jun-Wei Hsieh
• Format: Others
• Language: en_US
• Published: 2005
• Online Access: http://ndltd.ncl.edu.tw/handle/60398832849773793516

碩士 === 元智大學 === 電機工程學系 === 93 === Video Compression is good for reducing transmission data but very vulnerable to different transmission noise or errors like package loss. In this thesis, a novel data-embedding H.264/AVC coding scheme for error resilience is proposed to conceal and protect various transmission errors. This scheme uses two complementary methods to tackle and reduce the effect of transmission errors. First of all, the encoder uses a technique of data embedding to embed important information to the intra-coded (I) frame and the inter-coded (P) one. The important data includes the intra-coded type and motion vector which are extracted from the intra-coded frame and the inter-coded frame, respectively. When transmitting, the important data is embedded into the AC-coefficients of the H.264/AVC bitstream. Once the video stream is transmitted, the decoder will immediately extract desired embedded data from the stream for error recovering if a transmission error is detected. If unfortunately the embedded data can not be well extracted, the decoder will use the second method to conceal the effect of this error as can as possibly. The second method takes advantages of the BNM (Best neighborhood matching) scheme to find an optimal motion vector for video decoding. Since a smaller window mask is used and different weights are incorporated for balancing the contributions of neighbor windows, the task of correspondence search can be achieved very efficiently and effectively. Even though the corrupted video stream has higher packet loss rate (up to 20%), a higher quality of H.264/AVC video sequence still can be well recovered and maintained. Experimental results have proved the superiority of the proposed method in the efficiency, effectiveness, and robustness of error concealment and resilience for video coding.