Color Invariant Object and Texture Recognition

A fast, color-based algorithm for unknown illuminant is presented. of the object's color distribution R, G and B denote the 3 color 1: recognizing objects and textures viewed under an Objects are indexed by six numbers: the angles and the angles of the color edge distributions. If )ands of the image of an object (stretched out as vectors) then the color angular index comprises the 3 inter-band angles (one per pair of vectors). Similarly the edge angular index comprises the angles of within band color edges represented as feature vectors. In the general case the distribution of colors, and in turn the angular index, will depend on the color of the illuminant. If, however, the original color bands are transformed by a sharpening transform before computing the distribution angles, then the angular index is illuminant independent. The sharpening transform in this case refers to linear combinations of the sensor responses which yield narrow band or sharper sensors. Indexing using angles calculated post sharpening delivers excellent recognition for a variety of illuminations. Textures are represented by the six spatial correlation functions of the intra and inter band image color distributions. Treating these correlations as vectors, a new set of angular invariants is computed as the set of angles between the six vectors. The spatial correlation angles are shown to preserve illumination independence under the sharpening transform. Tests show that these angles possess sufficient discrimination to correctly identify a set of textures. The color and edge angular index is used to represent the spatial interactions in color textures and provides a rich representation for color textures. A coarse form of texture classification and segmentation is developed by computing these texture features locally across the image.