- From: Rik Cabanier <cabanier@gmail.com>
- Date: Sun, 15 Jul 2012 19:22:31 -0700
- To: Kai-Uwe Behrmann <ku.b@gmx.de>
- Cc: Calculemus <calculemus1988@gmail.com>, www-svg@w3.org
- Message-ID: <CAGN7qDCR-bzYgUh_+Y9Me=p9337L2CMR67cUEOnQkxCa8Ujfkg@mail.gmail.com>
On Sat, Jul 14, 2012 at 1:14 AM, Kai-Uwe Behrmann <ku.b@gmx.de> wrote: > Rik Cabanier <cabanier@gmail.com> schrieb: > > >On Fri, Jul 13, 2012 at 10:36 AM, Calculemus > ><calculemus1988@gmail.com>wrote: > > > >> About HDR, the problem is they can take values bigger than 1. > >> Let's say I have two HDR images with values in range 0-4. > >> Do I first map the range 0-4 to 0-1, do the blend math, and go > >> back to 0-4? Do I just do the math on the range 0-4 and > >> then clamp? > >> > > Clamping should always be the last step to preserve the HDR information > as long as possible during processing. HDR highlights behave other than > flat LDR white while compositing. > I will ask around at Adobe. I know that our static art application such as Illustrator and Photoshop generally clip after blending but I have heard that it might not be the case in the video editing world (ie After Effects might not do this.) > > >When you blending 2 images, you need to make sure that they are in the > >same > >colorspace (aka the blending colorspace). > >If they are, I think you can just apply the blending formula's to the > >raw > >values. (1 in the blending formulas represents the maximum value of the > >your colorspace). > >If they are not, or the colorspace has no maximum/minimum values, you > >will > >need to convert to a wide gamut colorspace, blend in that and then > >convert > >back. > > You can of course us a logarythmic function or a tone mapping op, which > looks much better than per channel clipping. The later one easily causes > not so nice colour casts. > > If your colorspace is wide enough there should not be any clipping. Make sure it is a wide gamut RGB if you want the same blending. Blending in CMYK or Lab looks very different for the same blend mode.
Received on Monday, 16 July 2012 02:22:59 UTC