[Webkit-unassigned] [Bug 220972] New: Color Gamut for VP8 and VP9 seems wrong

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=220972

            Bug ID: 220972
           Summary: Color Gamut for VP8 and VP9 seems wrong
           Product: WebKit
           Version: Safari 14
          Hardware: Unspecified
                OS: Unspecified
            Status: NEW
          Severity: Normal
          Priority: P2
         Component: WebRTC
          Assignee: webkit-unassigned at lists.webkit.org
          Reporter: agouaillard at gmail.com
                CC: youennf at gmail.com

Problem:
Safari and iOS incorrectly interprets video levels with VP8/9 encoded content.

My understanding is that most UVC devices as well as Chrome and OBS publish methods by default use “partial” video levels. (Partial is OBS terminology, This is similar to the “legal” vs “Full Range" broadcast video levels). Almost all UVC/Webcams that I’ve used transmit with partial video levels by default. This is also true of hardware encoders. Chrome and Firefox both assume and interpret partial video levels, thus displaying a correct looking image. Safari and iOS for some reason expects and interprets video as Full Range levels. This means that blacks appear dark grey instead of truly black, and whites appear light grey instead of truly white. This has a significantly negative effect on the image accuracy and quality.

This is not a problem with H.264. Safari/iOS correctly interprets video as “partial” and looks correct and similar to Chrome.

I've found this issue is present in Millicast, Google Meet, and Jitsi Meet. It appears to be an issue specifically with how Safari/iOS interprets video that’s encoded in VP8/VP9.

Here is a link to a test video clip with a greyscale ramp. Also in this folder are two screenshots from a waveform monitor showing the difference between Chrome output and Safari output
https://f.io/qdncjeuk

Let me know if you have any questions.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.webkit.org/pipermail/webkit-unassigned/attachments/20210126/0fde46c3/attachment.htm>