Perhaps for the blending operations that are clearly nonsensical for HDR
values, the blend function should clamp the inputs to [0,1]. E.g., better to
treat values > 1 as equivalent to 1.0, rather than suddenly get negative
results.
-- lg
On Jul 11, 2012, at 10:28 PM, Will Rosecrans wrote:
> Meh, I don't think screen mode really makes sense in FP anyway. It is
> meant is a hack to simulate what adding should do when you aren't
> working with linear images. It was always a bit of a weird 8 bit
> hack, and it breaks counterintuitively with anything outside the 0-1
> range. You expect that screening a bright thing with a bright thing
> should be very bright...
>
> a = .5, b = .5
> result = (.5+.5) - (.5*.5) = .75
> result is brighter than either input, makes sense.
>
>
> a=1.0, b=1.0
> result = (1 + 1) - (1*1) = 1
> "maximum brightness"
>
>
> a = 4.0, b = 4.0
> result = (4 + 4) - (4*4) = -8
> WTF, it turned negative?!
>
>
> I don't know that it makes sense to implement screen when we are
> assuming linear, potentially HDR images. Once you have everything
> linear, you just use an add to get the effect of adding two things.
>
>
> On Wed, Jul 11, 2012 at 10:19 PM, Stefan Stavrev
> <[email protected]> wrote:
>> You are fast! :D
>>
>> Ok I think I can do this for the other operations.
>>
>> Just to share another example for the blending mode "Screen",
>> whose formula is: f(a,b) = (a+b) - (a*b).
>>
> _______________________________________________
> 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
