OK, I think I get it. Let's call the feature "lum=1" instead of "rgb=1", since the important distinction is not that it's an RGB image (it could be RGB and adjust contrast separately for each channel, as is the default), but rather that you want to stretch the luminance.
I *think* that contrast should only adjust color channels, not alpha. Anybody have a contrary opinion? Or should it leave alpha alone by default, but have an option that means "stretch the alpha, too?" On Jul 25, 2012, at 12:18 AM, Stefan Stavrev wrote: > Larry, > > 1. "If using luminance leads to better results, should we be doing that > by default, and only do it for each channel separately if some other > flag is set (perchan=1)?" > > I see two options. > > Option 1: this is what I was going for, apply contrast to all > channels separately, even for 3-channel images. It does not > produce that bad results for RGB images. Example: > > http://postimage.org/image/d5d0v6y6d/ > http://postimage.org/image/irqldypvl/ > http://postimage.org/image/aer0niy41/ > http://postimage.org/image/o7n5qqykl/ > > But when it does produce a bad result, the user can tell us it is > RGB image and we can try a better solution, that is apply > contrast to the luminance channel for the RGB image. > > Option 2: Apply contrast separately to all channels if the > image has channels different than 3, and if it has 3 then > assume it is RGB and apply contrast to the luminance channel. > If the image has 3 channels, but it is not RGB, then you can > say perchan=1 and contrast will be applied to all channels. > > 2. "And I'm assuming that contrast leaves any designated alpha > channels untouched? " > > I wanted to ask you what to do about alpha and Z channels. Really not > sure how to handle this stuff, we will have to deal with it for all > the other operations too. That is why I wanted to solve channels > masks early on. > > 3. "Can the contrast be a color as well as a float? For example, can you > oiiotool in.png --contrast 1.5,1.5,0.8,1 -o out.png > and would that add contrast to R and G, reduce it a bit in B, and leave > the alpha channel alone?" > > Yep that is what those numbers would do. This is a very good idea, > to be able to set contrast per channel. > > > _______________________________________________ > 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
