Correcting the textures

The above image is the result of tone correcting the render output from the renderer. Although the scene looks naturally illuminated, the overall appearance is somewhat washed out and still lacks contrast. The main reason for the washed out appearance in this scene is because the textures were prepared without taking tone correction into consideration.

Let's recall two facts: 1) Since the avent of the sRGB standard, the standard monitor is assumed to have a gamma of 2.2. 2) All digital photos are now tone corrected with the equivalent of a gamma 2.2 so they can be viewed correctly on the standard monitor. Those two facts have important hidden consequences for the 3D CG artist.

The first most important consequence related to fact #1, is that because computer monitors are setup so they are suitable for looking at gamma corrected photos, this means that they are not suitable for looking at 3D CG textures.

The second most important consequence related to fact #2 is that because digital photos are now tone corrected, this means that they are suitable for viewing on a computer monitor but they are not suitable for making 3D CG textures.

Fortunately, there are solutions to this dilemma. First, most graphics and paint applications, today, have a color management setup that allows the artist to customize a viewing environment suitable for viewing, processing and making 3D CG textures. Second, photos planed to be used for 3D CG textures may be shot in RAW mode or at least, for stock photos, it is relatively easy to remove the tone correction so it can be used for texturing.

Before going into the procedural explanation of those two solution, it is important to understand why digital photos are unsuitable for using them as a basis for 3D CG textures unless they are preprocessed to remove the sRGB tone mapping.

A 3D CG renderer is an engine that computes light bouncing off of object surfaces. The light bouncing calculations may be very simple as in a raytracing or a scanline/z-buffer engine, or they may be much more complex and realistic as in a radiosity or GI engine. Whatever the render engine complexity, though, they all base their calculations on the reflectance of all the surfaces in the scene.

The reflectance of a surface indicates the spectral absorption and reflection characteristics of a surface. It is the color of the surface. If a surface reflects back 90% of the reds and absorbs 90% (in other words, it only reflects 10% back) of the blues, greens and yellows, then the surface color will look bright saturated red. In the RGB world where 3D CG artists live, we are only concerned with the red, green and blue absorption factors.

Reflectance is actually a more complex phenomenon than explained above but I will not go into those details for now. I will only mention that true reflectance is characterized by its Bidirectional Reflectance Distribution Function, known as its BRDF. Instead of holding only one RGB value per pixel, a BRDF, holds a 6-dimensional array of RGB values per pixel and one of those RGB values is selected based on both the camera position and the light position relative to the surface. For compactness, the 6D array of RGB values is actually stored by using a mathematical model instead of storing all the individual values.

So, in order to create realistic textures, it is necessary to encode, in each pixels of the texture map, the RGB reflectance. The digital camera, actually records this reflectance. This is the data that is recorded by the camera image sensor. Recall that each camera image sensor pixel counts (so to speak) the number of photons that it receives and converts that number of photons into a voltage.

The voltage of each sensor pixel is proportional to the number of photons received. And the number of photons received is proportional to the reflectance of the portion of the surface that is captured by the sensor pixel. However, because a tone correction is applied to the digital photo, this reflectance information is now all warped into something that is more suitable for viewing but does not hold the relative reflectance information anymore. In a typical tone corrected photo, a reflectance of 10% would turn out 35% which is quite a lot more than the reality.

Getting the right reflectance on 3D CG surfaces is most important when used with a Global Illumination renderer. If the textures are coming directly from tone corrected photos, which increase the brightness of the photos, the GI scene will look way too washed out. This is due to the fact that in a GI renderer, indirect light is calculated from reflectance characteristics of all the surfaces in the scene. Brighter reflectance will mean more indirect light bounced back into the scene than would be the case in reality. It is also important to use the right reflectance with raytracing or scanline renderer.