Fast Inter Prediction for H.264 to HEVC Transcoding

Abstract

H.264 video compression standard has achieved great success in real life applications. It has been widely used for video streaming and video storage. Even Bluray Disc (BD) adopts H.264 as its specifications. However, the next generation of video compression standard HEVC is completed in 2013. Under the same subjective quality, HEVC can save about 50% bitrate compared with H.264 [1]. Thus, Abstr act. Owing to the high compression efficiency improvement of High Efficiency Video Coding (HEVC) standard, it is expected as a successor to H.264. The widely usage of H.264 today and the expected future adoption of HEVC raises a new demand for H.264 to HEVC transcoding. An efficient transcoding way is necessary. In this paper, the proposed pixel domain transcoding architecture reuses the coding information such as motion vector, prediction mode and residual coefficient extracted from the H.264 decoder. An adaptive search range algorithm for motion estimation and several CU/PU pruning strategies for mode decision are proposed to speed up the encoder. The experimental results show that our proposed transcoding scheme can maintain the coding performance and save about 40% encoding time.Keywor ds: H.264, HEVC, transcoding, mode decision, adaptive search range. Intr oductionWith the improvement of video coding technology, storage capacity, display resolutions, and CPU processing capabilities, the applications of multimedia systems become wider and more popular. Therefore, how to efficiently provide video contents to users under different constraints is very important. Video transcoding is one of the efficient solutions to this problem. The transcoding technique provides video adaption in terms of bitrate reduction, resolution reduction, and format conversion.

Efficient H.264-to-HEVC Transcoding Based on Motion Propagation and Post-Order Traversal of Coding Tree Units

“…However, other factors may motivate the use of motion refinement. For example, Zong et al observed that motion refinement improves the coding efficiency when the H.264 MV and the AMVP selected by the HEVC encoder are different [8].…”

Fast H.264 to HEVC transcoder based on post-order traversal of quadtree structure

“…Several fast mode decision approaches have been proposed in the literature [3][4][5][6][7][8][9][10][11][12]. For example, Fang et al use H.264 modes, encoded residual and variance of motion vectors (MVs) to skip some prediction unit (PU) modes [3].…”

“…Shen et al select the H.264's dominant MV as a starting point, and refine this MV with a search range of 4 pixels [10]. Other authors propose to use a dynamic search range [3,8]. For example, Zong et al use the advanced motion vector prediction (AMVP) as a starting point, and define the search range as the distance between this point and the H.264's dominant MV [8].…”