With the recent discussions related to gamma handling I thought I'd post 
an image comparing how transmit behaves with respect to assumed_gamma in 
a texture overlaid upon an image base texture. The behavior of transmit 
at an assumed_gamma (ag) of 1.0 is something I like less well than its 
behavior at an ag of 2.2.

In the attached image I am comparing ag of 1.0 to an ag of 2.2 in each row.

There is a base image_map texture on a plane. Over the top of that 
texture is another with 5 columns of 4 colors (0,0.5,1.0,2.0) where the 
5 columns have transmit values of: 0,0.5,1.0,1.5,2.0. One parallel light 
source, camera orthographic, ambient 0, diffuse 1.0.

The non-linear behavior of transmit in the ag 2.2 case is often helpful 
to the final appearance of layered textures - the gray out region is 
tighter. When trying to map a 2.2 texture say from Norbert's collection 
into my ag 1.0 working space, I start by converting any rgb colors to 
srgb (bottom row), but that is usually not enough if the transmit value 
in any part of the texture is not 0.0 or 1.0.

I usually end up thrashing around with colors and transmit values to get 
closer, but truth is I'm often not able to match exactly the ag 2.2 
look. Might be different if I started from scratch I guess.

Am I missing some technique? Are there proposals in the works - or 
better available - which might help?

Aside: I have been wondering for a while about a more probabilistic 
pixel based transmit mode for blending overlapping textures. Might such 
an approach help with the dulling/graying of overlapping partly 
transparent textures...

Bill P.

Post a reply to this message

Attachments:
Download 'comparetransmitgammahandling.jpg' (876 KB)

Preview of image 'comparetransmitgammahandling.jpg'

Am 18.10.2016 um 18:29 schrieb William F Pokorny:

> The non-linear behavior of transmit in the ag 2.2 case is often helpful
> to the final appearance of layered textures - the gray out region is
> tighter. When trying to map a 2.2 texture say from Norbert's collection
> into my ag 1.0 working space, I start by converting any rgb colors to
> srgb (bottom row), but that is usually not enough if the transmit value
> in any part of the texture is not 0.0 or 1.0.

Your example uses very unsaturated colours; to get the full picture, you
should also try the same experiment with highly saturated colours. I can
imagine that you'll come to different conclusions there.


> Aside: I have been wondering for a while about a more probabilistic
> pixel based transmit mode for blending overlapping textures. Might such
> an approach help with the dulling/graying of overlapping partly
> transparent textures...

What do you mean by "probabilistic pixel based transmit mode"?

Post a reply to this message

>> Aside: I have been wondering for a while about a more probabilistic
>> pixel based transmit mode for blending overlapping textures. Might such
>> an approach help with the dulling/graying of overlapping partly
>> transparent textures...
>
> What do you mean by "probabilistic pixel based transmit mode"?

Quantum simulation of photons :-)

Post a reply to this message

On 10/18/2016 02:36 PM, clipka wrote:
> Am 18.10.2016 um 18:29 schrieb William F Pokorny:
>
>> The non-linear behavior of transmit in the ag 2.2 case is often helpful
>> to the final appearance of layered textures - the gray out region is
>> tighter. When trying to map a 2.2 texture say from Norbert's collection
>> into my ag 1.0 working space, I start by converting any rgb colors to
>> srgb (bottom row), but that is usually not enough if the transmit value
>> in any part of the texture is not 0.0 or 1.0.
>
> Your example uses very unsaturated colours; to get the full picture, you
> should also try the same experiment with highly saturated colours. I can
> imagine that you'll come to different conclusions there.
>

Will put it on the todo list - would suppose I'd notice the greying 
transitions less in such an experiment. The image used was one where I'd 
struggled with transparent textures and getting a "blend" between 
textures that I wanted.

>
>> Aside: I have been wondering for a while about a more probabilistic
>> pixel based transmit mode for blending overlapping textures. Might such
>> an approach help with the dulling/graying of overlapping partly
>> transparent textures...
>
> What do you mean by "probabilistic pixel based transmit mode"?
>
I have in mind some sort of sampling between textures weighted by the 
transmit values involved. Suppose somewhat like the blending which 
happened in the tilt shift image I posted recently. I admit though, I'm 
thinking aloud.

Bill P.

Post a reply to this message

On 10/19/2016 04:11 AM, scott wrote:
>>> Aside: I have been wondering for a while about a more probabilistic
>>> pixel based transmit mode for blending overlapping textures. Might such
>>> an approach help with the dulling/graying of overlapping partly
>>> transparent textures...
>>
>> What do you mean by "probabilistic pixel based transmit mode"?
>
> Quantum simulation of photons :-)
>
:-) With those new quantum computers, it will be easy.

Post a reply to this message

On 18-10-2016 18:29, William F Pokorny wrote:
> [snip]
> There is a base image_map texture on a plane. Over the top of that
> texture is another with 5 columns of 4 colors (0,0.5,1.0,2.0) where the
> 5 columns have transmit values of: 0,0.5,1.0,1.5,2.0. One parallel light
> source, camera orthographic, ambient 0, diffuse 1.0.
>

Out of curiosity: what is a transmit value >1.0?

-- 
Thomas

Post a reply to this message

On 10/19/2016 07:16 AM, Thomas de Groot wrote:
> On 18-10-2016 18:29, William F Pokorny wrote:
>> [snip]
>> There is a base image_map texture on a plane. Over the top of that
>> texture is another with 5 columns of 4 colors (0,0.5,1.0,2.0) where the
>> 5 columns have transmit values of: 0,0.5,1.0,1.5,2.0. One parallel light
>> source, camera orthographic, ambient 0, diffuse 1.0.
>>
>
> Out of curiosity: what is a transmit value >1.0?
>
People use >1 sometimes to amplify colors or increase contrast.

I think Norbert used this technique for one of his recent, 
http://www.tc-rtc.co.uk/, metal monster images.

It isn't the most stable technique as can be seen in the right part of 
the posted images. I'd say use values in the 0 to 1 range as a rule.

Aside: I have a question rattling around in my head as to why the higher 
values tend toward black.

Bill P.

Post a reply to this message

> With the recent discussions related to gamma handling I thought I'd post
> an image comparing how transmit behaves with respect to assumed_gamma in
> a texture overlaid upon an image base texture. The behavior of transmit
> at an assumed_gamma (ag) of 1.0 is something I like less well than its
> behavior at an ag of 2.2.

Is the problem here though is that you're trying to solve this by 
tweaking gamma? Maybe it works for this issue, but it's also going to 
affect all the other lighting and colours in your scene.

The better approach would be to figure out how to achieve what you want 
without changing the scene-wide gamma setting. What if you adjust the 
gamma of the image and the "gamma" of your rgbt values (ie don't step in 
equal steps for each column)? You might have to sit down and work out 
the maths to get what you want without changing assumed_gamma.

Post a reply to this message

Am 19.10.2016 um 12:21 schrieb William F Pokorny:
> On 10/19/2016 04:11 AM, scott wrote:
>>>> Aside: I have been wondering for a while about a more probabilistic
>>>> pixel based transmit mode for blending overlapping textures. Might such
>>>> an approach help with the dulling/graying of overlapping partly
>>>> transparent textures...
>>>
>>> What do you mean by "probabilistic pixel based transmit mode"?
>>
>> Quantum simulation of photons :-)
>>
> :-) With those new quantum computers, it will be easy.

Such a simulator already exists -- it's called Reality ;)

Post a reply to this message

Am 19.10.2016 um 12:20 schrieb William F Pokorny:

>>> Aside: I have been wondering for a while about a more probabilistic
>>> pixel based transmit mode for blending overlapping textures. Might such
>>> an approach help with the dulling/graying of overlapping partly
>>> transparent textures...
>>
>> What do you mean by "probabilistic pixel based transmit mode"?
>>
> I have in mind some sort of sampling between textures weighted by the
> transmit values involved. Suppose somewhat like the blending which
> happened in the tilt shift image I posted recently. I admit though, I'm
> thinking aloud.

You mean, something like using the "average" pseudo-pattern to textures,
but with the weight being modulated by the textures' "transmit" channel?

While that sounds interesting, a major drawback would be that the
"transmit" channel couldn't be used for its original purpose, namely
specifying transparency of the texture.

Post a reply to this message