The color and texture of the surface pigmentation pattern of skin images provide important diagnostic and prognostic features for detection of lethal skin-cancer. Since intensity as well as texture are important in this application, the segmentation algorithm is required to incorporate both of them in the process of region extraction and description. A multichannel segmentation scheme is presented here, in which the intensity-based segmentation is obtained using the modified pyramid-based region extraction algorithm while the texture-based segmentation is obtained by a bi-level shifted-window processing algorithm that uses new generalized cooccurrence matrices. The results of each channel, based on individual segmentations, are then analyzed using heuristic rules to obtain the final color and texture based segmentation. The extracted regions are then described through a multilayered neural net for a model-based classification for predicting malignancy.