Temporal transcoding of H.264/AVC video to the scalable format

Abstract

In this work, a novel implementation of a video transcoder that converts a video sequence encoded with the H.264/AVC standard to a temporally scalable H.264/SVC stream is achieved with the use of a pixel-domain heterogeneous architecture. The input H.264/AVC stream is fully decoded by the transcoder. Macroblock coding modes are extracted from the input stream and are reused to encode the output stream. A set of new motion vectors is computed from the input stream coded motion vectors, and are mapped to either the hierarchical B-frame or zero-delay referencing structures employed by H.264/SVC. These new motion vectors are further subjected to a 3 pixel refinement. As a result, a significant decrease in computational complexity is achieved, while maintaining a close to optimum compression efficiency.