Color Angular Indexing

A fast color-based algorithm for recognizing colorful objects and colored textures is presented. Objects and textures are represented by just six numbers. Let r, g and b denote the 3 color bands 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 image vectors). The color edge angular index is calculated from the image's color edge map (the Laplacian of the color bands) in a similar way. These angles capture important low-order statistical information about the color and edge distributions and invariant to the spectral power distribution of the scene illuminant. The 6 illumination-invariant angles provide the basis for angular indexing into a database of objects or textures and has been tested on both Swain's database of color objects which were all taken under the same illuminant and Healey and Wang's database of color textures which were taken under several different illuminants. Color angular indexing yields excellent recognition rates for both data sets.