Computer vision and image processing software tool with application to color segmentation for skin cancer diagnosis

Today's scientists and engineers require sophisticated data visualization and analysis tools to help them understand, discover and communicate information about the phenomena they study. The need for a powerful and comprehensive computer vision and image processing tool inspired the development of CVIP tools (Computer Vision and Image Processing tools). The environment consists of a central set of libraries that provide for an application programmer interface (API) for researchers to develop their own applications, and also a program that provides a central user interface and ties in the functionality of the API. An application to color segmentation for skin cancer diagnosis using the software was developed. The six segmentation algorithms explored are the fuzzy c-means, spherical coordinates transform/center split, principal coordinates transform/median cut, split and merge, adaptive thresholding and multiresolution segmentation algorithms. The algorithms are utilized as part of an overall border finding process that includes noise removal, morphological data reduction, object localization and contour encoding. The algorithm was applied to 66 skin tumor images selected by a dermatologist as a representative set. The best results were obtained with the border finding process used in conjunction with the pct/median cut segmentation algorithm.