That's another possibility I would entertain -- automatically upgrading the 
subsampling when a sufficiently high quality is requested.

So, just to be explicit, are you proposing making these changes yourself and 
submitting a patch? I'm basically on board with this general design idea.

I set the default quality for jpeg output (if no "quality" attribute is found) 
quite high, 98. I'm also happy to revisit whether that's too much. I can't say 
it was a principled reason for that particular value.



On Oct 16, 2014, at 6:39 PM, Justin Israel <[email protected]> wrote:

> Awesome.
> I would say it would seem natural to stick with the existing behaviour for 
> the default, which follows your philosophy of the reasonable defaults. And 
> for those in my situation where they need to control the sampling, they can 
> specifically apply the setting. FWIW, it appears the stated Imagemagick 
> behaviour is to use 4:4:4 automatically if the quality value is >= 90, and 
> disregard any sampling settings that were passed (or the default). 
> 
> On Fri, Oct 17, 2014 at 1:17 PM, Larry Gritz <[email protected]> wrote:
> Yes, sounds reasonable. Let's set/accept an attribute called 
> "jpeg:subsampling" set to one of those values for explicit control.
> 
> We want to detect and set it in the reader as well, so that "copy"-like 
> operations from jpeg to jpeg preserve the sampling of the original unless 
> overridden.
> 
> Now we just need to haggle over the proper default.
> 
> http://en.wikipedia.org/wiki/Chroma_subsampling
> This states fairly clearly that 4:2:0 (what we do now) is the usual for 
> JPEG/JFIF, MJPEG, DVD, BluRay, H.264, and many others. That's probably why 
> the jpeg library uses this default. Maybe Nuke and ImageMagick go out of 
> their way to request a somewhat higher quality? I'm tempted to argue for 
> keeping the default as it is, and with the extra control, those who care 
> could request 4:2:2 or 4:4:4. But I'm willing to be swayed if consensus is 
> that the default should be higher quality.
> 
> 
> 
> On Oct 16, 2014, at 2:59 PM, Justin Israel <[email protected]> wrote:
> 
>> So maybe the solution is to offer some constants like: 
>> "4:4:4" - Calls jpeg_set_colorspace() with JCS_RGB
>> "4:2:2" - does this, but uses SET_COMP(0, 1, 2,1, 0, 0,0); for the first 
>> component instead
>> "4:2:0" - Does exactly what it does now
>> "4:1:1" - uses 4,1  1,1  1,1 for the 3 components
>> Does that sound reasonable? Then you get as much control as Nuke/imagemagick 
>> would offer 
>> 
> 
> --
> Larry Gritz
> [email protected]
> 
> 
> 
> 
> _______________________________________________
> Oiio-dev mailing list
> [email protected]
> http://lists.openimageio.org/listinfo.cgi/oiio-dev-openimageio.org
> 
> 
> _______________________________________________
> Oiio-dev mailing list
> [email protected]
> http://lists.openimageio.org/listinfo.cgi/oiio-dev-openimageio.org

--
Larry Gritz
[email protected]



_______________________________________________
Oiio-dev mailing list
[email protected]
http://lists.openimageio.org/listinfo.cgi/oiio-dev-openimageio.org

Reply via email to