Submacroblock motion compensation for fast down-scale transcoding of compressed video

摘要: For downscale (spatial resolution reduction) video transcoding, previous work has shown that motion vectors in the input can be reused so costly estimation process avoided. Since derived may not perfectly aligned with ones, a drift compensation loop consisting of inverse transform and (or domain compensation) modules is required. However, always performed based on macroblocks full resolution, which makes full-resolution inevitable. This fact renders as next computational bottleneck. Through analytical modeling transcoding architectures, we propose new architecture employees submacroblock method. method enables to entirely reduced resolution. In addition, memory space required for significantly since frame buffers are no longer Extensive analysis experiments show proposed system achieves further reduction computing load comparing prior systems minimum quality degradation.