In computer vision, there are several methods to create a 3-D model of an object. One of them, called “photometric stereo,” uses multiple images of the object under different light directions. In this 3-D reconstruction technique, we need to calculate surface normals of the object as an intermediate step, and this picture is the result of that step.
The image depicts the surface normals of a mirrored sphere. The “surface normal” is the direction perpendicular to a surface, and in this visualization different normal directions are represented with colors. For example, red means the object surface is facing to the right, green means the object surface is facing upwards, and blue means the object surface is facing towards the viewer, and other colors are combinations of these directions.
The surface normals depicted in this image are not all geometrically correct because the algorithm assumes that the surface is not shiny like a mirror, so what we see in this image are actually some artifacts caused by highlights and shadows.
– Sema Berkiten