Fast partitioning decision making for prediction units on H.264-to-HEVC transcoding using machine learning
Yan Soares, Guilherme Corrêa, Luciano Volcan Agostini · 2019
With the high computational complexity of transcoding video from the H.264/AVC standard to the state-of-the-art High Efficiency Video Coding (HEVC) standard, new approaches must be explored to fasten up the adaptation of all legacy content. This work proposes a fast decision algorithm to reduce the transcoding complexity between the H.264/AVC and HEVC video standards using partitioning information from the H.264/AVC macroblocks to fasten up the partitioning decisions of Prediciton Units (PUs) on the HEVC reencoding process. This strategy allowed a 25% reduction on transcoding time with a compression efficiency loss of just 0.745% in comparison with the original transcoder.