Linked by Thom Holwerda on Mon 25th Mar 2013 21:09 UTC
Legal Late last week, Nokia dropped what many consider to be a bomb on the WebM project: a list of patents that VP8 supposedly infringes in the form of an IETF IPR declaration. The list has made the rounds around the web, often reported as proof that VP8 infringes upon Nokia's patents. All this stuff rang a bell. Haven't we been here before? Yup, we have, with another open source codec called Opus. Qualcomm and Huawei made the same claims as Nokia did, but they turned out to be complete bogus. As it turns out, this is standard practice in the dirty business of the patent licensing industry.
Thread beginning with comment 556820
To view parent comment, click here.
To read all comments associated with this story, please click here.
RE[11]: Big picture...
by saynte on Wed 27th Mar 2013 12:27 UTC in reply to "RE[10]: Big picture..."
saynte
Member since:
2007-12-10

Yeah, I noticed it was PSNR as well, which I had heard (although I am not directly knowledgeable about it) that if one optimizes strictly for it, it can prefer some "blurriness" in the output image. I thought SSIM had been proposed to compensate for that deficiency(?).

In any case, I think we generally conclude that we have at least two options (x264 and the VP8 encoders) which have both acceptable output.

Play time over, back to work for me ;) .

Reply Parent Score: 2

RE[12]: Big picture...
by lemur2 on Thu 28th Mar 2013 05:45 in reply to "RE[11]: Big picture..."
lemur2 Member since:
2007-02-17

Yeah, I noticed it was PSNR as well, which I had heard (although I am not directly knowledgeable about it) that if one optimizes strictly for it, it can prefer some "blurriness" in the output image. I thought SSIM had been proposed to compensate for that deficiency(?).


Actually, I believe it is the reverse. VP8 tends to blur areas of "high motion", whereas h.264 tends to add small "artefacts" (features which were not present in the uncompressed data). AFAIK, PSNR just objectively compares the still frame(s) of the uncompressed image with the image after compression, and so larger "areas of blur" are "marked harshly" under PSNR compared with smaller "artefacts".

The thing is that when looking at a video played at normal speed, the human eye tends to see areas of high motion as blur anyway. So subjectively the as-rendered VP8-compressed video can look better to people than the as-rendered h.264-compressed video, even though the latter can have a slightly better rating as measured by PSNR.

I think SSIM may well have been proposed to compensate for the deficiency in PSNR for penalising areas of blur where there was high motion. If that is the case, then SSIM would give a better rating to VP8 than it gets under PSNR. Giving comparison figures in PSNR actualy under-rates VP8, AFAIK.

PS: On investigation, I think PEVQ (not SSIM) might be the comparison method that attempts to take account of the way that people actually perceive the playing video stream.

http://en.wikipedia.org/wiki/PEVQ

In any case, I think we generally conclude that we have at least two options (x264 and the VP8 encoders) which have both acceptable output.


Agreed. They both have acceptable output. One of them has far more acceptable terms of use, especially for a web standard.

Edited 2013-03-28 06:00 UTC

Reply Parent Score: 2

RE[13]: Big picture...
by saynte on Thu 28th Mar 2013 07:25 in reply to "RE[12]: Big picture..."
saynte Member since:
2007-12-10

I think all quantitative measures share the same flaw, that they do not necessarily relate to real perceived quality, although SSIM was designed to relate better to how people perceive image quality (I don't know if that's true in practice).

Reply Parent Score: 2