An expert subjective evaluation of a Reshaper model vs HDR10 for HDR coding

Abstract

This paper presents the results of a subjective evaluation experiment, made to compare different HDR coding technologies, conducted at the recent ITU/ISO/IEC VCEG/ MPEG/JPEG Meeting in San Diego, CA, February 2016. A set of “anchor” streams, conforming to the HDR10 spec, was compared to a similar rate-matched set obtained using a method called “Reshaper”, which requires normative changes to the underlying HEVC Main10 Profile. The subjective evaluation experiment was done involving sixteen “video coding expert” subjects, sitting two-at-a-time in front of a 4000-nit capable SIM2 monitor, at a distance equal to 2.5 times the screen height. The test itself followed the Expert Viewing Protocol (EVP) developed by one of the authors (V. B.). EVP involves the sequential viewing of (a) uncompressed original, (b) codec A, (c) codec B. To minimize as much as possible judgment biases, several randomizations in the presentation order were used. Performance results are presented in a graphical form for ease of analysis. It is noted that the Reshaper model provides some visual benefits in a minority of cases, notably at a mid-rate setting, while differences at lower and higher rates were within a margin of error. This paper focuses on the test methodology and results more than the specifics of the technology under test.