DocumentCode
729585
Title
Which metric can predict coding gain of H.265/HEVC over H.264/AVC?
Author
Jianhua Xiao ; Li Song ; Zhengyi Luo ; Rong Xie ; Wenjun Zhang
Author_Institution
Inst. of Image Commun. & Network Eng., Shanghai Jiao Tong Univ., Shanghai, China
fYear
2015
fDate
17-19 June 2015
Firstpage
1
Lastpage
5
Abstract
Subjective evaluation conducted by JCT-VC (Joint Collaborative Team on Video Coding) members shows that H.265/HEVC achieves about 50% rate saving over H.264/AVC without sacrificing subjective quality. In this paper, we study 13 objective image and video quality assessment (IQA/VQA) metrics, including the recently proposed ones - FSIM [1], GMSD [2] and IWSSIM [3], in terms of coding gain prediction of HEVC over H.264. Experimental results on HEVC Class B and Class C test sequences show that most of the metrics underestimate the rate saving. Surprisingly, a relatively old metric - Noise Quality Measure (NQM) [4] index consists well with subjective evaluation. To verify the universality of the phenomenon, we carried out further tests on another ten video sequences with different levels of spatial and temporal complexities. The experimental results show that NQM still predicts rate saving more accurately than the other metrics.
Keywords
video coding; FSIM; GMSD; H.265-HEVC; HEVC Class B test sequences; HEVC Class C test sequences; IQA metrics; IWSSIM; JCT-VC members; NQM; VQA metrics; coding gain prediction; image quality assessment; joint collaborative team on video coding members; noise quality measure; rate saving; spatial complexities; subjective evaluation; subjective quality; temporal complexities; video quality assessment; video sequences; Complexity theory; Encoding; Image quality; Indexes; Measurement; Quality assessment; Video sequences; BD-Rate; HEVC; objective quality assessment;
fLanguage
English
Publisher
ieee
Conference_Titel
Broadband Multimedia Systems and Broadcasting (BMSB), 2015 IEEE International Symposium on
Conference_Location
Ghent
Type
conf
DOI
10.1109/BMSB.2015.7177228
Filename
7177228
Link To Document