Tandent - Intrinsic Imaging

Each pixel in an image is generated as a function of the color of its material, and the color(s) of the light source(s) at that position. Therefore, varying the color or position of the light sources can significantly change the appearance of an object. State-of-the-art Computer Vision techniques often either ignore such illumination effects, try to reduce its influence by e.g. extracting features based solely on gradient information, or augment their datasets by including multiple images of the same object taken under different lighting conditions.

At Tandent Vision Science, our goal was to automatically segregate the material color (reflectance) from the light source (illumination). The significance of this technology is vast, and can potentially have a profound impact on the field of Computer Vision as a whole. One example of such a segregation is given below.