Improving the parallelization efficiency of HEVC decoding

Abstract

In this paper we present a new parallelization approach for HEVC decoding called Overlapped Wavefront (OWF). It is based on wavefront processing and improves its parallelization efficiency by allowing overlapped execution of consecutive pictures. Furthermore, in this strategy of the decoding steps are performed on a CTB basis rather than on a picture basis, which improves data locality. Our implementation achieves between 29.6%, 42.4%, and 66.6% higher frame rates compared to previous results and 11.3%, 21.0%, and 38.0% higher frame rates compared to Tiles, for 2160p, 1600p, and 1080p, respectively.

Saliency-Guided Complexity Control for HEVC Decoding

“…Souza et al [17] achieved the HEVC decoding acceleration, which benefits from the parallel computing of Graphics Processing Unit (GPU). Similarly, [25] presented a new parallelization approach for accelerating HEVC decoding speed with higher arXiv:1610.02516v5 [cs.MM] 9 Jan 2018 frame rate. The above approaches can save HEVC decoding time in some specific hardware, but they cannot reduce the complexity and power consumed by HEVC decoding.…”

Parallel HEVC Decoding on Multi- and Many-core Architectures

“…The scalability of wavefront processing is limited by the reduced number of independent blocks (CTBs or macroblocks) at the beginning and at the end of each frame. To solve this limitation, and increase the parallel scalability of WPP, a technique called Overlapped Wavefront (OWF) has been proposed [8]. With OWF multiple pictures can be decoded simultaneously resulting in a more constant parallelism during execution.…”

HEVC performance and complexity for 4K video

“…First approaches to enable real-time decoding of HEVC coded 4K video sequences have been analyzed and presented in [7], [8], [9], and [10]. In these studies, the HEVC test model (HM) reference software decoder code was optimized and modified to support multithreading.…”

“…Entropy slices are not part of the final standard but with Wavefront Parallel Processing (WPP), similar multi-threaded decoding as with entropy slices can be achieved. A slightly modified version of WPP, called Overlapped Wavefront (OWF) and Tiles have been studied in [8] and [9] based on HM 4.1. The most recent publication shows results for OWF based on HM 8.0 and further reports speedup due to the use of Single Instruction Multiple Data (SIMD) code optimizations [10].…”