Efficient Rate Control Mechanism and Its Applications Based on Region-of-Interest Video Coding

• Main Authors: Ming-Chieh Chi, 紀明傑
• Other Authors: Mei-Juan Chen
• Format: Others
• Language: en_US
• Published: 2008
• Online Access: http://ndltd.ncl.edu.tw/handle/20231260434625902638

博士 === 國立東華大學 === 電機工程學系 === 97 === In a variety of multimedia applications, the storage and transmission of image and video always deeply attract the users. However, since the data transmission has to meet the constraint of channels with time-varying bandwidth, a flexible and adaptive rate control mechanism is needed to handle with the corresponding parameters in the encoder and maintain the limited bandwidth during transmission. In low bitrate video transmission, the region-of-interest (ROI) video coding based on the rate control mechanism provides a solution to enhance the visual quality in the region-of-interest area. Besides being used in the applications of the ROI video coding, the rate control mechanism also plays a critical role in the encoder when the transmission bandwidth is limited. In this dissertation, the region-of-interest determination is first proposed to select the ROI area. And then, the region-of-interest video coding algorithm based on the video coding standards is further proposed to enhance the visual quality in the ROI area through the rate control mechanisms. In the rate control algorithm, the quantization parameter initialization, the bit allocation algorithm for the frame layer and the quantization parameter determination in the macroblock layer are proposed to improve the coding efficiency. In this dissertation, two ROI video coding frameworks are proposed in H. 263+ and H. 264 video coding standards. In H. 263+, a skin color extraction is used to select the ROI regions and the fuzzy logic is employed in the proposed ROI video coding algorithm to arrange the importance of the macroblocks and then calculate a suitable quantization parameter to the current macroblock. Through attention models based on visual rhythm, a fast ROI determination is proposed to obtain the ROI region by using four sampling lines in the consecutive frames. By combining the visual rhythm ROI determination with the skin color extraction, a lookup table is established to indicate the importance of each macroblock. In H. 264, the flexible macroblock ordering (FMO) functionality is further used to calculate the corresponding quantization parameter. Experimental results show that the proposed ROI determination can obtain the ROI effectively and efficiently, and the proposed ROI video coding algorithms can significantly enhance the quality of ROI regions in both H. 263+ and H. 264. For the rate control algorithm in H. 264+, a quantization parameter initialization is proposed to give a suitable quantization parameter to the first I frame by jointly considering the target bitrate and the video content. In the frame layer, the proposed rate control algorithm adaptively adjusts the target bitrate according to the information of neighboring encoded frames. The information is collected from a sliding window to show the characteristics of these frames. In the macroblock layer, the target bitrate is estimated by the information of neighboring encoded macroblocks. Experimental results show that the proposed algorithms can improve the coding efficiency.