Humans are able to judge the shape of an undulating surface from variations in the amount of light reflected from it due to changes in its orientation/position relative to the light source: shape-from-shading. The study of shape-from-shading in humans has a long history but remains an open problem. While it is clear that humans can use shading information as a shape/depth cue the mechanisms which support this process are not well understood. This knowledge gap is also found in the computer vision literature where despite many years’ of effort and the existence of many good algorithms to deal with special cases, a robust, generic solution to shape-from-shading remains somewhat elusive. In this chapter, we review the principle findings from the literature on shape-from-shading including aspects of lighting estimation and the role of prior assumptions.
|Name||Advances in Computer Vision and Pattern Recognition|