Modified IDCT kernel for down-conversion of MPEG-2 compressed video

A traditional method for down conversion of MPEG-2 video is to perform low-pass filtering and sub-sampling after full decompression. However, this method has two shortcomings; large memory requirement and high computational complexity. To overcome these drawbacks, down conversion in the DCT domain has been preferred. Previous methods need only a quarter of the full frame memory because completely reduced frames are stored in the frame memory. But the loss of field information often causes significant errors during motion compensation. To avoid this problem, recent schemes adopt a hybrid structure where field information is preserved by performing down conversion in the DCT domain only for the horizontal direction, and down conversion in the spatial domain for the vertical direction. In this paper, we propose a novel technique for the horizontal down conversion in the DCT domain where a modified 4-IDCT kernel is adopted, and apply it to the hybrid down conversion decoder. The proposed 4-IDCT kernel outperforms the conventional 4-IDCT kernel in that it performs efficient low-pass filtering and sub-sampling concurrently. We apply the proposed technique to down conversion of MPEG-2 compressed videos. Experimental results show that the proposed scheme provides high visual quality while maintaining reasonable complexity.