LOSSLESS ENHANCEMENT CODING FOR H.264/AVC SCALABLE EXTENSION(International Workshop on Advanced Image Technology 2007)

概要

論文の詳細を見る
Based on our seminal investigation of the H.264/AVC 4x4 integer transform, which discovered quite interesting coefficient properties, this paper introduces an efficient lossless scalable coding scheme with a base layer of Joint Scalable Video Model (JSVM), a state-of-the-art reference software for the emerging scalable coding extension of the H.264/AVC standard. To quantitatively compare the enhancement coding performance of our proposal against that of the conventional pixel-wise residual method, we measure the entropy of pixel-wise difference between the base layer video and original video. The comparison confirms that our method is far superior to the pixel-wise method. We also conduct coding experiments using various frame sizes, frame rates, GOP structures, and base quantization parameter (QP) values. Starting with a JSVM compatible base layer, overall (=base+enhancement) lossless compression rates of 54% (with intra-frame base layer coding) compared to the original file size and 44% (with inter-frame base layer coding) are attained by averaging the results collected from 14 sequences. Our intra-frame coding performance is as efficient as Motion JPEG2000 lossless mode, while our inter-frame coding offers 12% more compression. Moreover, insensitivity of overall performance to QP selection in the base layer, which is a desirable feature, is demonstrated.
社団法人電子情報通信学会の論文
2007-01-03