Sign in to follow this  
lipsryme

Problem with Color Correction and 3D LUT

Recommended Posts

lipsryme    1522

I'm having some issues with color correction using a 3D LUT.

I'm using a neutral LUT 16x16x16 from the UnrealEngine page here: http://udn.epicgames.com/Three/rsrc/Three/ColorGrading/RGBTable16x1.png

 

I've open it up in photoshop and saved it as a Volume Texture DDS (32bpp unsigned ARGB8) using the nvidia plugin.

 

And here's my shader:

float4 PS(in VSOutput input) : SV_Target
{
	float3 rawColor = InputTexture0.Sample(LinearClampSampler, input.UV).rgb;

	float3 scale = (float3(16, 16, 16) - 1.0f) / float3(16, 16, 16);
	float3 offset = 1.0f / (2.0f - float3(16, 16, 16));

	float3 colorCorrected = LUT.Sample(LinearClampSampler, scale * rawColor + offset).rgb;
	
	return float4(colorCorrected, 1.0f);
}

The code is from this sample here: http://http.developer.nvidia.com/GPUGems2/gpugems2_chapter24.html

Can someone spot the problem ? : /

 

Not sure if the problem has to do with the order of rendering.

I'm currently tonemapping my linear result to a R8G8B8A8_UNORM_SRGB target, then use this as the input for the color correction which then outputs to my R8G8B8A8_UNORM_SRGB backbuffer.

 

Here's a screen of what my result looks like:

http://d.pr/i/xGYR

 

And here's the original for comparison:

http://d.pr/i/I6SL

Edited by lipsryme

Share this post


Link to post
Share on other sites
kauna    2922
float3 Scale = (16.0f - 1.0f) / 16.0f; //volume texture size
float3 Offset = 1.0f / (2.0f * 16.0f);
 
 
Here's my code for the scale and offset calculations. Seems that your offset has a slight error.
 
Cheers!

Share this post


Link to post
Share on other sites
kauna    2922

You are probably sampling off the pixel center? Either use point sampling or load instruction or correct your sampling position.

 

Cheers!

Share this post


Link to post
Share on other sites
lipsryme    1522

I'm using DX11 so I it shouldn't...

When using point sampling it looks horrible with a lot of banding artifacts (also doesn't fix the pixel borders):

http://d.pr/i/4dP7

 

How would I need to correct it to work ?

Edited by lipsryme

Share this post


Link to post
Share on other sites
kauna    2922

That's really strange ... It should work with point sampling too, and it shouldn't produce any banding artifacts.

 

Are you using MSAA ?

 

Cheers!

Share this post


Link to post
Share on other sites
lipsryme    1522

I'm not...
Do you suppose this is related to gamma correction issues? I just don't see what could possibly wrong. I'm using the output of my tonemapping as the input scene which is in regular 32 bit unsigned format and being output to an srgb target after being linear.

 

update: Gamma correction does not seem to be the problem. I've tried using regular R8G8B8A8_UNORM for both outputs. Besides being in the wrong color space the border issue was still there. Maybe it's the LUT texture ?

 

update2: I just tried making it the only post effect in my chain, and I've tried using regular LDR for shading. The pixel border still appears. This has to be either an issue with the input texture or the sampling process..

If I just output rawColor inside the color correction shader I get the perfect original scene. But using the LUT I get the border.

It looks almost like an edge highlighting. Using LinearWrap makes it even more apparent.

Edited by lipsryme

Share this post


Link to post
Share on other sites
kauna    2922

Only difference I can see between your and my implementation that I use SampleLevel(s_linear,Color,0) to sample the volume texture, so that mip maps won't affect the output.

 

[edit]Are you able to use "load" instruction? It should be equivalent to point sampling, although you'll index the texture instead of using texture coordinates.

 

Cheers!

Edited by kauna

Share this post


Link to post
Share on other sites
kauna    2922

Glad to be able to help. I was fighting the same code (with different issues though) last week.

 

Cheers!

Share this post


Link to post
Share on other sites
kauna    2922

Curious, something to do with gamma correction somewhere? Is your volume texture format defined as _SRGB or not? 

 

Cheers!

Share this post


Link to post
Share on other sites
lipsryme    1522

It's not I'm doing gamma conversion in shader during lookups.

Though sampling it as linear makes it too dark.

 

Edit: Is there an order I have to maintain when I apply the color transforms in photoshop ?

Edited by lipsryme

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

Sign in to follow this