This paper examines texture as a surface property which in conjunction with the imaging geometry and illumination creates the observable texture in the image. To use texture as a cue to image understanding, one must be able to use the observed image texture in order to recover texture characteristics that are invariant to imaging and illumination conditions. Two major approaches are discussed for this purpose: detailed surface reconstruction using photometric stereo, and illumination invariant feature definition for sets of pixels with particular relevance in the latter case to remote sensing applications.
展开▼