Adaptive Early Termination Algorithm Using Coding Unit Depth History in HEVC

• Published in: Journal of signal processing systems Vol. 91; no. 8; pp. 863 - 873
• Main Authors: Bae, Jong Hyun, Sunwoo, Myung Hoon
• Format: Journal Article
• Language: English
• Published: New York Springer US 15.08.2019; Springer Nature B.V
• Subjects: Adaptive algorithms; Algorithms; Circuits and Systems; Coding; Computer Imaging; Delay; Electrical Engineering; Engineering; Image Processing and Computer Vision; Model testing; Pattern Recognition; Pattern Recognition and Graphics; Signal,Image and Speech Processing; Standard deviation; Video compression; Vision; Fast CU decision; Video compression; Inter prediction; Temporal correlation; High efficiency video coding
• ISSN: 1939-8018; 1939-8115
• DOI: 10.1007/s11265-018-1399-y

This study proposes an adaptive early termination algorithm using a coding unit (CU) depth history (AECH). AECH uses the temporal correlation of co-located coding tree units (CTUs) and traces a CU depth history. For tracing the CU depth history, AECH computes the average of all CU depth values and finds the standard deviation of the CU depth history from the average. The target CU depth range is then adaptively chosen based on the standard deviation from the mean. The proposed algorithm can skip the CU depths, which are out of range. Thus, it can considerably reduce computational complexity. AECH yields time-saving (TS) performance of 42.15 and 41.08% (low delay B) compared to the high efficiency video coding test model (HM) 9.0 and 12.0 version, respectively. Furthermore, AECH shows 40.51% TS (low delay P) in HM 12.0. In addition, the proposed algorithm increments Bjontegaard delta bitrate (BDBR) at only approximately 1.22 and 1.04% (low delay B) in HM 9.0 and HM 12.0, respectively. Moreover, AECH achieves 1.26% BDBR (low delay P) in HM 12.0.