Breaking the H.264 spec for a new Codec - H265?

[iwod]

Where is H265? to give a new higher quality codec that guarantee minimum quality. The current issues are everything is described as Mpeg 4. When the world ask for Mpeg 4 their perception was Mpeg 4 is higher quality. However ASP is Mpeg 4, AVC is also Mpeg 4, and the difference between Baseline Profile and High Profile, are also Mpeg 4.

Mpeg 4 in itself no longer gives satisfactory results.   It is one thing i like Divx, when it states a devices play Divx HD, you know it support High Profile etc. Currently there are many mobile devices out there that does not support H.264 High Profile or support what ever that describe as Mpeg 4. It is a Mess!

With x264 already mature, What happen if DS break the h.264 standard and try to forge a new format based on h.264 / x264 with better features like replace the i16x16 mode with an actual 16x16 transform and make the inter transform work accordingly (16x16, 8x8, or 4x4). Maybe make the decision quadtree instead of per-MB and get energy-preserving quantizer?

Making a new Codec from everything we have learned with x264 without the restriction of a H.264 spec.

[Dark Shikari]

A practical problem with arbitrarily modifying the H.264 spec (with the intent of actually creating something seriously widespread, as opposed to just a toy) is that for large-scale commercial adoption, you need to deal with the patent problem.  At least if I recall correctly, the MPEG-LA license only covers actual H.264 implementations, not things-which-rip-off-parts-of-H.264, so as far as I know you can't get a license for something that isn't quite H.264.

Of course, as a project it would be quite interesting nonetheless.

[kierank]

At least if I recall correctly, the MPEG-LA license only covers actual H.264 implementations, not things-which-rip-off-parts-of-H.264, so as far as I know you can't get a license for something that isn't quite H.264.

RealNetworks seem to have an H.264 licence but produce a ripped-off version.

[Dark Shikari]

RealNetworks seem to have an H.264 licence but produce a ripped-off version.

AFAIK Real licenses patents individually (or at least claims to).

[iwod]

So, let say if an entity license the needed patents, we could theoretically build an much improved x264 ( or what ever it is called ) without restriction?

[Biggiesized]

How much better would x264 be if the developers had access to the MPEG LA patent portfolio, especially Panasonic's?

[Dark Shikari]

How much better would x264 be if the developers had access to the MPEG LA patent portfolio, especially Panasonic's?

I don't know of any patented encoder techniques that would be useful to x264.  In my experience, most patents are extremely specific and thus relatively easy to work around.  Most are also hardware-oriented.

Of course, I explicitly don't read such patents.  They're a waste of time and even less informative than academic papers.

[iwod]

So how much more can we squeeze out, with the current spec.

Lookahead quant and trellis ME, psy improvement, iterative intra coding, multi-level b-pyramid, better weightp.

What sort of improvement if we have all these in?

Sorry for Noob question, Would these work on Basic Profile?

[weasel]

http://www.h265.net/2011/01/hevc-test-model-hm-reference-software-1-0-released.html