Creating PBR maps out of several different images with varying lighting can be a great way to get accurate normal- and displacement maps - but the method also comes with some limitations.
How does it generally work?
The camera is placed on a tripod above a surface, all outside lighting is reduced (closing blinds, turning off all lights, etc.) and the area is instead lit from only one side. This process gets repeated 4 or 8 times with the light source rotating around the area in steps of 90° or 45° respectively. The resulting images are then combined using software to create a fully de-lit color map and a normal map which can be used as a basis for creating the rest of the material.
The final result is a material with much more accurate displacement than what could be achieved with standard Bitmap Approximation.
When can it be used?
In short: This method is helpful whenever you want to capture a small area with subtle displacement.
Examples of this are individual leaves or small pieces of fabric. One big challenge when creating these scans lies in the fact that the center of the frame will always almost have a better, more accurate displacement than the corners.
This makes it a more attractive solution for individual objects (like leaves) than full seamless surfaces (though these are still possible, you just have to crop a bit more into the center).
A camera. I usually use my Nikon Z50 but a smartphone camera can certainly be used as well.
A tripod (can’t be a monopod as it has to stand perfectly still.)
- A remote control for the camera (optional, but recommended)
A directional light source that you can easily move around and aim in different directions. I use a light wand.
A plain background. This can be a piece of paper but if you want to go the extra mile you can use a light table to also capture translucency.
- Substance Designer
- An image editor
It’s important for the success of this method that the camera and the object in question don’t move during the entire capturing process, even subtle changes can create artefacts.
Having a leaf that is in the middle of becoming limp will introduce unwanted movement to the shot. Leaves are also usually not entirely flat. For that reason I like to add two pieces of tape to the back of every leaf. This ensures that it stays flat on the surface.
The shooting process
Place your object on a plain background to which you have access from all sides. Mount the camera on a tripod and point it downwards towards the object. Place the light source next to the target area and point it towards it. It doesn’t matter with what angle you start as long as it is a multiple of 45° or 90° from the camera’s perspective (I always start lighting from the left/west). Use this opportunity to set up the exposure and focus of the camera beforehand. You want to press as little buttons as possible after taking the first shot.
Once everything is configured properly, take the first shot (using the remote if you have one). From now own you must not move your camera or your subject at all until you have taken the last photo. Move your light source by 45° or 90° around the target area and take the next picture. Repeat this until you have a set of 8 or 4 images showing the target area in specific lighting conditions. If you are using a light table you can turn it on afterwards and capture an additional image with just the backlight shining through.
The light staff that I use for multi-angle approximation.
Recording the “South-West” lighting angle of an individual leaf. The lighting pad remains turned off for this process.
Recording the additional “scattering” with the light pad turned on.
Here is the final set of pictures for one leaf. This includes the 8 lighting angles and the additional “scattering” image. All images were captured without touching either the leaf or the camera to ensure that there is no movement between them.
|Additional “subsurface” image with the backlight turned on.|
Editing in Substance Designer
Create a new graph in Substance Designer and load your 4 or 8 images into it. Then add a “Multi-Angle to Albedo” and a “Multi-Angle to Normal” node. Select the appropriate amount of inputs in both node’s settings. For the normal node you also need to select the lighting direction in the first image as well as the direction that you rotated your light source. The albedo node only needs the different images and has no further settings.
These two nodes will give you a normal- and a color output which is a great starting point for all the other maps. I usually continue by feeding the normal map into a “Normal to Height HQ” node which generates an displacement map out of the normal map. Generating the displacement out of the normal map is not 100% accurate but it works pretty well for organic shapes like leaves.
The roughness map can be created using a “Curvature (Smooth)” node together with a “Levels” node (for fine tuning).