Arbitrary ROI-Based Wavelet Video Coding

An arbitrary shape region of interest (ROI) coding is presented for scalable wavelet video codec in this paper. The framework of ROI-based wavelet coding is shown in the Figure 1. First, by segmenting and tracking of the original video, we can obtain the region of interest for each frame. The shapes information of the ROI is given in the way of two-value template (MASK). The ROI can be one or more of video object. The padding technology of macroblock in reference frame is employed in each ROI to fill the boundary macroblocks with some samples in ROI and transparent macroblocks without the samples in ROI. The motions are estimated by using hierachial variable size block matching. According to the motion vectors, the pixels in current macroblocks can be mapped to the reference samples. The samples outside the ROI are re-assigned the values of motion vectors. The motion compensated temporal filtering (MCTF) can be performed along the motion trajectory using LG5/3 or other lifting filters. At the end of the first stage, the low-frequency frames are input into the second-level filtering, while high-frequency frames are output. The MASKs and the positions of transparent macroblocks are delivered in the way of odd frames in accordance with low-frequency ones, even frames corresponding to high-frequency ones.