| Summary: | 碩士 === 國立中正大學 === 資訊工程研究所 === 83 ===
The motion compensated discrete cosine transform coding(MCDCT) is an efficient image sequence coding technique. The motion compensated technique can remove the temporal redundancy between the current frame and the previous reconstructed frame. The spatial redundancy of the motion compensated predicted errors can be removed using the discrete consine transform (DCT). For reducing the bit rate, we can classify the blocks into the moving or nonmoving. Only the moving blocks have to be encoded using the motion compensated DCT. That is, the motion vector and the predicted error for the moving blocks are transmitted to the decoder. The nonmoving blocks are not encoded and predicted by the same blocks in the previous frame. This method can be called the adaptive motion compensated discrete consine transform coding. (AMCDCT).
In AMCDCT, each block needs one more bit for the moving information. Hence, we propose a new adaptive quadtree motion compensated discrete consine transform coding (AQMCDCT) to further reduce the bit rate. In our proposed AQMCDCT algorithm, the overhead information is encoded by a quadtree structure. Each leaf of the quadtree represents the moving information of a variable size block. As will to be shown, the improvement of AQMCDCT over the AMCDCT is up to 0.71 dB at nearly the same bit rate.
Im order to further reduce the bit rate for the quantizied DCT coefficients and keep the visual quality, we propose an adaptive edge-based quadtree motion compensated desecrate cosine transform (AEBQMCDCT) By a 2-step selection scheme, the nonedge blocks will be encoded at lower bit rate but the edge blocks will be encoded at higher bit rate. Hence, the edge information is preserved and the image quality is improved. The edge information is not only used for classifying the edge blocks and the nonedge blocks, but also used for selecting the important DCT coefficients that will be encoded. That is, only part of the DCT coefficients, not all the DCT coefficient, is encoded to reduce the number of encoded DCT coefficients. The edge blocks will be further classified into four different classes according to the orientaion and location of the edges. Each class of edge blocks selects the different set of the DCT coefficients to be encoded. By this method, we can just preserve and encode a few DCT coefficients, but still maintain the visual quality of the images. In the proposed AEBQMCDCT image sequence coding scheme, the average bit rate of each frame is reduced to 0.072 bit/pixel and the average PSNR value is 32.11 dB.
|
|---|
