A simple and fast multi-channel filtering algorithm is presented in th
is paper for texture edge detection. The algorithm requires only a sma
ll number of channels and automatically determines the important chann
el parameters. It does not assume a priori knowledge about the type an
d number of textures occurring in the input image. It operates in an e
ssentially unsupervised mode and involves only limited human intervent
ion. The algorithm is tested extensively with a variety of synthetic a
nd natural textures. Despite the simplicity of the overall algorithm,
the segmentation results obtained are encouraging.