3 ms·

HEVC over Twitter

29 points by bratao 13y ago

garblegarble 13y ago
That's pretty neat, although it's unfortunate they don't explain what's causing the edges around the right side of the face, I assume it's some interesting adaptive macroblock division - any HEVC-wise people know?
Danieru 13y ago
"[HEVC is] comparable to DLI" I'm not sure what images they're looking at but from my end they are not comparable. Not at all. You cannot just pretend the giant mosaic tiling is not there. My annoyance is of course at the meta. Right now those with a shoe in the game are gearing up to push H.265. Remember a year or so back during the fight over H.264 in web video? Some commentators pointed out that the patents covering H.264 will expire in due time. That will never happen. Proprietary versus open source is not H.264 versus WebM. Both sides will innovate and improve and if we do not make an effort towards open source it will not fall into our lap. The proprietary side is taking those hundreds of millions in ill-gotten licensing fees and pouring it back into the next generation so they can continue to seek rent. Meanwhile those fighting for our rights are getting along on shoe-string budgets and buckets of passion. If we want an open future we need to support those who have our best interests in mind.
- rayiner 13y ago
  > The proprietary side is taking those hundreds of millions in ill-gotten licensing fees and pouring it back into the next generation so they can continue to seek rent. Meanwhile those fighting for our rights are getting along on shoe-string budgets and buckets of passion. Rent-seeking is defined as manipulating the political environment to achieve economic gain without creating value. The process you yourself describe isn't at all rent-seeking. If the end-result, H.265, is better than open source alternatives, then clearly value is being created. What you're describing, in your own words, is the use of licensing to fund capital-intensive research that cannot be adequately performed without capital investment (on "shoe-string budgets."). Your central complaint seems to be that MPEG/VCEG isn't giving away the fruits of their capital-intensive research for free. Well I hate to break it to you, but neither is Google (WebM). Indeed, I think its far better what MPEG is doing: simply taking money in return for a license, as opposed to what Google is doing, which is subsidizing WebM development with their advertising empire.
  - pgeorgi 13y ago
    If MPEG-LA and their collaborators were simply taking money in return for a license, all would be well. They're not. "Vorbis|Theora|VP8|Opus|... is infringing on tons of patents, with no indemnification" is an infamous part of their routine, even though they never point to patents nor provide indemnification themselves (which is impossible when it comes to patents). With such FUD in place, it's not about being "better than open source alternatives" anymore. For that they can go die in a fire.
- 0x09 13y ago
  This is a good thread to bring this particular terminology into, because when I followed the links in the post to DLI to find out about the algorithm, I was greeted with a page containing no details, no specification, and links to a Windows .DLL and EXE.* In contrast, if I want to learn about the other codec in this comparison, I can download Rec H.265 from the ITU freely, ISO draft HEVC specifications freely, final ISO documents for little more than the cost to keep ISO running, use an official BSD-licensed reference implementation, and already find several high-quality independent open source implementations (for a one year old standard). A standard which, by the way, was developed by not one but two international standards bodies, the proceedings of which I can view online (http://mpeg.chiariglione.org/meetings http://mpeg.chiariglione.org/meetings). Maybe I've been in this for too long, but when I started, codecs like DLI were the norm rather than the exception. The top competitors at the time were Sorenson, On2 (VPx), RealNetworks, and Microsoft, and MPEG-4 had barely trickled out of the demo boxes in the form of a cracked copy of a MS draft implementation that ended in ";-)". Patent dangers were well known thanks to Fraunhofer's actions against LAME, but since they didn't extend to Europe where much open source codec development occurred, and since they were (before GPLv3) largely within a separate legal sphere from copyright issues, and since it had long been folk wisdom that doing video compression without stepping on someone's patent was impossible, they were seen as little more than an unavoidable detail (here's some source code, compile it yourself in America.) It was none other than H.264 (ostensibly with a lot of help from x264) that put an end to this era. So when I see folks putting the MPEG Licensing Authority (unaffiliated with ISO and MPEG, by the way) on a pedestal, and I see it a lot, and using it as an excuse to further technologically inferior codecs which are themselves (unlike MPEG standards) technically proprietary, I have a hard time reconciling the thought process. And to bring it up in response to a comparison against something like DLI? Give me a break. * https://sites.google.com/site/dlimagecomp https://sites.google.com/site/dlimagecomp
  - ZeroGravitas 13y ago
    MPEG and MPEG-LA may not be formally affiliated but they're two sides of the same coin. If MPEG continue to create codecs without any attempt to limit patents (and associated royalties and control) included in them then MPEG-LA or similar will continue to exist, and be made up of basically the same corporations. Not that its a big conspiracy, just a mistake that's taken on a self-sustaining, and self-funding, life of its own. Relatedly, remember when MPEG told everyone that H.264 would have a royalty free baseline? http://www.robglidden.com/about-2/why-a-royalty-free-mpeg-a-manifesto/ http://www.robglidden.com/about-2/why-a-royalty-free-mpeg-a-...
mherdeg 13y ago
FYI, you can fit about 4370 bits in a Tweet ( https://blogs.oracle.com/ksplice/entry/1st_international_longest_tweet_contest https://blogs.oracle.com/ksplice/entry/1st_international_lon... , https://blogs.oracle.com/ksplice/entry/the_1st_international_longest_tweet https://blogs.oracle.com/ksplice/entry/the_1st_international... ).
aidenn0 13y ago
I can't be the only one who thought that the DLI image was significantly better than the HEVC image (see the last set of 3 images in the article)?
0x09 13y ago
I thought it would be interesting to see how transmitting only the 251 (byte-size) highest-energy coefficients in the entire image would fare, which is somewhat of an idealization of the method used by the codecs in this comparison, but the result is not visually as good as either: http://i.imgur.com/TJaMs0f.png http://i.imgur.com/TJaMs0f.png This is a cheap imitation of an encoder: there is no entropy coding, but neither is there consideration for scan order, so there are some biases for and against. Assuming an entropy coder could reduce the bits to 67% there's a little improvement: http://i.imgur.com/ILY23uM.png http://i.imgur.com/ILY23uM.png Comparing them, it seems the block partitioning and filtering employed by DLI and H.265 become much more perceptually important than the raw mathematical effectiveness of the transform under bit starved rates like this.

29 points by bratao 13y ago

That's pretty neat, although it's unfortunate they don't explain what's causing the edges around the right side of the face, I assume it's some interesting adaptive macroblock division - any HEVC-wise people know?

Danieru 13y ago

"[HEVC is] comparable to DLI" I'm not sure what images they're looking at but from my end they are not comparable. Not at all. You cannot just pretend the giant mosaic tiling is not there. My annoyance is of course at the meta. Right now those with a shoe in the game are gearing up to push H.265. Remember a year or so back during the fight over H.264 in web video? Some commentators pointed out that the patents covering H.264 will expire in due time. That will never happen. Proprietary versus open source is not H.264 versus WebM. Both sides will innovate and improve and if we do not make an effort towards open source it will not fall into our lap. The proprietary side is taking those hundreds of millions in ill-gotten licensing fees and pouring it back into the next generation so they can continue to seek rent. Meanwhile those fighting for our rights are getting along on shoe-string budgets and buckets of passion. If we want an open future we need to support those who have our best interests in mind.

rayiner 13y ago
> The proprietary side is taking those hundreds of millions in ill-gotten licensing fees and pouring it back into the next generation so they can continue to seek rent. Meanwhile those fighting for our rights are getting along on shoe-string budgets and buckets of passion. Rent-seeking is defined as manipulating the political environment to achieve economic gain without creating value. The process you yourself describe isn't at all rent-seeking. If the end-result, H.265, is better than open source alternatives, then clearly value is being created. What you're describing, in your own words, is the use of licensing to fund capital-intensive research that cannot be adequately performed without capital investment (on "shoe-string budgets."). Your central complaint seems to be that MPEG/VCEG isn't giving away the fruits of their capital-intensive research for free. Well I hate to break it to you, but neither is Google (WebM). Indeed, I think its far better what MPEG is doing: simply taking money in return for a license, as opposed to what Google is doing, which is subsidizing WebM development with their advertising empire.
- pgeorgi 13y ago
  If MPEG-LA and their collaborators were simply taking money in return for a license, all would be well. They're not. "Vorbis|Theora|VP8|Opus|... is infringing on tons of patents, with no indemnification" is an infamous part of their routine, even though they never point to patents nor provide indemnification themselves (which is impossible when it comes to patents). With such FUD in place, it's not about being "better than open source alternatives" anymore. For that they can go die in a fire.
0x09 13y ago
This is a good thread to bring this particular terminology into, because when I followed the links in the post to DLI to find out about the algorithm, I was greeted with a page containing no details, no specification, and links to a Windows .DLL and EXE.* In contrast, if I want to learn about the other codec in this comparison, I can download Rec H.265 from the ITU freely, ISO draft HEVC specifications freely, final ISO documents for little more than the cost to keep ISO running, use an official BSD-licensed reference implementation, and already find several high-quality independent open source implementations (for a one year old standard). A standard which, by the way, was developed by not one but two international standards bodies, the proceedings of which I can view online (http://mpeg.chiariglione.org/meetings http://mpeg.chiariglione.org/meetings). Maybe I've been in this for too long, but when I started, codecs like DLI were the norm rather than the exception. The top competitors at the time were Sorenson, On2 (VPx), RealNetworks, and Microsoft, and MPEG-4 had barely trickled out of the demo boxes in the form of a cracked copy of a MS draft implementation that ended in ";-)". Patent dangers were well known thanks to Fraunhofer's actions against LAME, but since they didn't extend to Europe where much open source codec development occurred, and since they were (before GPLv3) largely within a separate legal sphere from copyright issues, and since it had long been folk wisdom that doing video compression without stepping on someone's patent was impossible, they were seen as little more than an unavoidable detail (here's some source code, compile it yourself in America.) It was none other than H.264 (ostensibly with a lot of help from x264) that put an end to this era. So when I see folks putting the MPEG Licensing Authority (unaffiliated with ISO and MPEG, by the way) on a pedestal, and I see it a lot, and using it as an excuse to further technologically inferior codecs which are themselves (unlike MPEG standards) technically proprietary, I have a hard time reconciling the thought process. And to bring it up in response to a comparison against something like DLI? Give me a break. * https://sites.google.com/site/dlimagecomp https://sites.google.com/site/dlimagecomp
- ZeroGravitas 13y ago
  MPEG and MPEG-LA may not be formally affiliated but they're two sides of the same coin. If MPEG continue to create codecs without any attempt to limit patents (and associated royalties and control) included in them then MPEG-LA or similar will continue to exist, and be made up of basically the same corporations. Not that its a big conspiracy, just a mistake that's taken on a self-sustaining, and self-funding, life of its own. Relatedly, remember when MPEG told everyone that H.264 would have a royalty free baseline? http://www.robglidden.com/about-2/why-a-royalty-free-mpeg-a-manifesto/ http://www.robglidden.com/about-2/why-a-royalty-free-mpeg-a-...

mherdeg 13y ago

FYI, you can fit about 4370 bits in a Tweet ( https://blogs.oracle.com/ksplice/entry/1st_international_longest_tweet_contest https://blogs.oracle.com/ksplice/entry/1st_international_lon... , https://blogs.oracle.com/ksplice/entry/the_1st_international_longest_tweet https://blogs.oracle.com/ksplice/entry/the_1st_international... ).

aidenn0 13y ago

I can't be the only one who thought that the DLI image was significantly better than the HEVC image (see the last set of 3 images in the article)?

0x09 13y ago

I thought it would be interesting to see how transmitting only the 251 (byte-size) highest-energy coefficients in the entire image would fare, which is somewhat of an idealization of the method used by the codecs in this comparison, but the result is not visually as good as either: http://i.imgur.com/TJaMs0f.png http://i.imgur.com/TJaMs0f.png This is a cheap imitation of an encoder: there is no entropy coding, but neither is there consideration for scan order, so there are some biases for and against. Assuming an entropy coder could reduce the bits to 67% there's a little improvement: http://i.imgur.com/ILY23uM.png http://i.imgur.com/ILY23uM.png Comparing them, it seems the block partitioning and filtering employed by DLI and H.265 become much more perceptually important than the raw mathematical effectiveness of the transform under bit starved rates like this.