Organisation internationale de normalisation

Project development, status, and guidance (3)

1. Communication to and by parent bodies (0)

See section 6.2.

2. Conformance test set development (0)

It was noted that many features are similar for SHVC and MV-HEVC, and that it would be desirable to test the same features in both cases. This could help ease the effort and increase the coverage of testing for conformance test set development for the two sets of extensions. See further notes in section 6.2.

3. Version 2 bug reports and cleanup (0)

See notes for S0003.

4. HEVC coding performance, implementation demonstrations and design analysis (2)

   1. Version 1 performance (1)

See also section 3.4.2 notes regarding verification testing.

13.0.0.1.1.1.1.1.61JCTVC-S0301 Some Results on Image Quality of HEVC and WebP [D. He (Blackberry), J. Wang (Magnum Semiconductor)] [late]

(Consideration of this topic was chaired by GJS on Thursday 10-23 p.m.)

(This was a late information document.)

This document reports some results on HEVC RExt as a still-image coding tool versus the "WebP" proprietary coding design, where the distortion is measured in both PSNR (peak signal-noise-ratio) and SSIM (structural similarity). On a set of grayscale images with resolutions from 512x512 to 2048x2560, it was reported that HM 16.1.1 consistently outperforms WebP in terms of rate-distortion performance, regardless whether the distortion measure is PSNR or SSIM. On average, at mid-quality range, the average PSNR BD-rate of HM 16.1.1 versus WebP (Method 6 = high-quality setting) was reported as −25.0%, and the SSIM BD-rate was reported as −16.6%; at high-quality range, the average PSNR BD-rate of HM 16.1.1 versus WebP (Method 6) was reported as −20.9%, and the SSIM BD-rate was reported as −13.7%.

It was remarked that it is tricky to consider a comparison of the performance of our design relative to third-party. It was also noted that the contribution was a very late submission. It was suggested that we should make sure to be clear that we are not endorsing a comparative assessment of a third-party technology under these circumstances.

The contribution reported closer correspondence between PSNR and SSIM behaviour at higher bit rates than at lower bit rates.

2.