Linear Workflow washes images out

[soulburn3d]

So obviously linear workflow is a better method mathematically to calculate the light in your renders. And you get results that are truer to your original textures and such. However, the side effect is your images get washed out (if you feel the term "washed out" has a negative connotation, feel free to replace it with the term "Low Contrast"). Here's a simple example...



both materials are the same grey level (the linear one is the grey with gamma set to 2.2), lights are adjusted to give the same grey value in the final render, the only difference between the images is one is linear and one is not. Notice the bottom image has way less contrast than the top image. If I run the linear image through a levels adjustment in photoshop, I can get back something closer to my original non linear image.

Here's a textured example, same issue...



Sometimes you want utmost mathematical realism, and sometimes you don't, so I was just wondering how people are generally dealing with this? Are people accepting the lack of contrast as the new normal? Are people using post processes to add contrast back? Are people adjusting their lights to get the most contrast they can? Just interested in hearing other people's experiences on the subject, I certainly have a number of techniques to solve the issue, I'm just wondering what other people are doing.

- Neil

[vlado]

So obviously linear workflow is a better method mathematically to calculate the light in your renders.

Nope. (The renderer already works in a linear physical space, there's nothing more to be done about it). Linear workflow is a method to a) input the proper color data into the renderer and b) view the results properly.

Sometimes you want utmost mathematical realism, and sometimes you don't, so I was just wondering how people are generally dealing with this?

I guess first of all, you need to adjust the gamma of the input textures. Also, if you set up a scene in linear workflow right from the start, you are automatically setting it up in a way so that the contrast is good (and then if you turn off the linear workflow you typically get a result that's way too dark).

Best regards,
Vlado

[FSGFX]

Neil here is a good explanation how this is working...
http://www.youtube.com/watch?v=D2iDv9hnQiw

[pixela]

Hi,

When you adjust gamma of the input textures + when you set your colors in LWF scene from the beginning, the contrast will be better.
Though it will not be as contrasted as gamma=1
My explanation will not be any scientific but I see LWF as adjusting the falloff curve for the transition from black to white. I preceive it as a falloff for greys so gamma=1 is a more steep transition but especially for interiors with small openings, LWF makes the greys look more similar to real life distribution of light.

[tricky]

I can't remember which scientist said it, but it goes something like this:

"If you think you understand quantum theory, you haven't understood it properly"

I think the same goes for Linear Workflow and all things 'gamma'.

Our pipeline uses some sort of pseudo linear workflow (though I have forgotten why we have things set up the way they are!). We just do a lot of painting/post production to get the images the way we want them. There is nothing particularly scientific. Definitely a lot more 'art' than physically correct-ness!

[soulburn3d]

I guess first of all, you need to adjust the gamma of the input textures. Also, if you set up a scene in linear workflow right from the start, you are automatically setting it up in a way so that the contrast is good (and then if you turn off the linear workflow you typically get a result that's way too dark).

Well, for my first example, I'm just using color and no texture, and then I set the scene up with linear workflow from the start, so I think the image you're referring to as "contrast is good" is the image I'm referring to as "washed out", and the image you're referring to as "way too dark" is what I consider "good contrast". There's no right and wrong here, it's an artistic choice (just like how many directors of photography will add contrast at the color grading stage, or pick a film stock known for its contrast), so I guess from what you're saying it sounds like to your eye the one I think is low contrast is what you consider to be normal. Thanks for the input.

Neil here is a good explanation how this is working...
http://www.youtube.com/watch?v=D2iDv9hnQiw

Thanks, I already know all of the theory behind linear workflow, this question is more about the artistic results, which doesn't seem to get discussed much.

I see LWF as adjusting the falloff curve for the transition from black to white. I preceive it as a falloff for greys so gamma=1 is a more steep transition but especially for interiors with small openings, LWF makes the greys look more similar to real life distribution of light.

Yup, I think that's an excellent way of looking at it, and that's exactly what I've observed. Maybe I'm just not interested in producing images that look exactly like real life

We just do a lot of painting/post production to get the images the way we want them. There is nothing particularly scientific. Definitely a lot more 'art' than physically correct-ness!

Thanks, so that's one vote for post processing.

- Neil

[joconnell]

Pretty much just s-curves for everything around here. Also Vlado is always mentioning that people often use the wrong brightness values for textures with whites being too whites and blacks being too dark - they might be contributing to too much light being bounced around a scene and washing things out a bit.

[soulburn3d]

I can certainly imagine that happening, although it's not the case in the images above. Looks like you're another vote for post processing, thanks for the feedback!

- Neil

[Morbid Angel]

I think, if you feel that the linear workflow washes out your image, I would recommend to adjust your lighting, rather then do a post processing work on the final render.

[soulburn3d]

So what lighting adjustments would you recommend? In the first example above, the lighting is a simple vray area light, I can make it darker, but that will darken the image evenly, which won't create more contrast. Would you recommend moving the light higher to add more shadows? Or in the case of an hdri providing Image Based Lighting, increase the contrast of the hdri to try and get contrastier lighting?

- Neil

[Morbid Angel]

Its quite simple, I think of linear workflow as just the fact that your lighting illuminates surface more, so you need less lights to get great looking result. Now with that said, what you are looking for is deeper shadow, like in the image you showed, what this would mean is playing around with your environment to get your lights to create the same feel as the non linear render. The shadow will always stay the same, no matter what method you use, its the surface shading that gets changed, the fill lighting has a big effect, I usually take the fill down quite a bit if I want to have deeper shadows.

[samuel_bubat]

We don´t use linear workflow at all. It´s a static solution to a dynamic problem. We rather use just reinhard color mapping. With it you can basically adjust your scene as you need it. With every image you may need other settings. with an outdoor scene you may want more contrast in an indoor scene you may need a much more extreme gamma curve to prevent burned out windows
and to dark interiors. With reinhard color mapping you can for example make a huge room with a tiny window absolutely bright without burning out the window area. Now depending on the settings you need to adjust the textures accordingly. Sometimes you need to increase the gamma (let´s rather say color curve because gamma actually only describes the middle grey value of an output device) sometimes you need to decrease it to prevent washed out textures. The whole linear workflow thing came up with the introduction of physical cameras wich act like real cameras and have no option for color curve adjustment.. because well these are the "handcuffs" a real camera has to deal with. In 3d and max you can come over these real world induced problems for quite a while (log. exposure for example). Sorry to say it again but i think linear workflow is utter nonsense

[soulburn3d]

Its quite simple, I think of linear workflow as just the fact that your lighting illuminates surface more, so you need less lights to get great looking result. Now with that said, what you are looking for is deeper shadow, like in the image you showed, what this would mean is playing around with your environment to get your lights to create the same feel as the non linear render. The shadow will always stay the same, no matter what method you use, its the surface shading that gets changed, the fill lighting has a big effect, I usually take the fill down quite a bit if I want to have deeper shadows.

Thanks for the suggestions. In the example I gave, there is no fill light or environment light, it's just a single vray area light, but reducing the fill is certainly a good idea for more complex lighting setups, which of course would be far more common than my simple example.

We don´t use linear workflow at all. It´s a static solution to a dynamic problem. We rather use just reinhard color mapping. With it you can basically adjust your scene as you need it. With every image you may need other settings. with an outdoor scene you may want more contrast in an indoor scene you may need a much more extreme gamma curve to prevent burned out windows and to dark interiors. With reinhard color mapping you can for example make a huge room with a tiny window absolutely bright without burning out the window area. Now depending on the settings you need to adjust the textures accordingly. Sometimes you need to increase the gamma (let´s rather say color curve because gamma actually only describes the middle grey value of an output device) sometimes you need to decrease it to prevent washed out textures. The whole linear workflow thing came up with the introduction of physical cameras wich act like real cameras and have no option for color curve adjustment.. because well these are the "handcuffs" a real camera has to deal with. In 3d and max you can come over these real world induced problems for quite a while (log. exposure for example). Sorry to say it again but i think linear workflow is utter nonsense

Thanks for that perspective. In some ways you're also saying "do post processing", except of course you're doing it in the render itself, which has a number of advantages over trying to manipulate the final pixel values.

I'm glad I'm getting plenty of varying responses here, it gives us all ideas when we decide for ourselves what's the method we want to use for our own workflow.

- Neil

[FSGFX]

i agree with all of you...
but if you have to integrate cg stuff into realworld footage... and using alle the raw passes to recreate the final image in nuke...
linear workflow is the easiest way to match the LUT from the camera...

if you do something else everything is a possible solution...

[Alan Iglesias]

Everything Samuel said...

-Alan