All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=600497

Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?keys=agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICgiIrvpAsM


Bot(s) for this bug's original alert(s):

chromium-webrtc-rel-linux

CL range:
http://test-results.appspot.com/revision_range?start=384572&end=384575

I see a WebRTC roll:
https://codereview.chromium.org/1850673003

There are 15 CLs in that roll, but nothing immediately stands out. nisse@, there are a couple CLs that look possibly related:
https://codereview.webrtc.org/1827023002
https://codereview.webrtc.org/1843413002

Can you take an initial look?

Issue 600515 has been merged into this issue.

Please note that other tests around the same range also alerted - see the graphs in https://chromeperf.appspot.com/group_report?bug_id=600515 (i.e. linked in duped bug 600515).

This decreases at the same time as  http://crbug.com/600503 , so it's very likely the issues are related.

Spoke with perkj@, and it does make sense that turning on the denoiser reduces video quality. Basically, it denoises the barcoded frame from the ref video, then encodes/decodes and captures the frame. However, when we do the PSNR comparison between the captured frame and the source frame we do not apply the denoising step to the source frame, so there will be a quality cost for the denoiser.

Since the root cause is the same as 600503 I'm duping this to that bug.