Integration of color and boundary information for improved region of interest identification in electron magnetic resonance images.

A Windows-based, object-oriented application system for segmentation and analysis of electron magnetic resonance (EMR) images is described. The integrated system is developed for better recognition of regions of interest (ROI) in murine EMR images. The system combines the clustering method of color segmentation with boundary detection, for efficient segmentation of regions of interest in EMR images. Initially, the red/green/blue (RGB) color space is converted into spherical coordinates transform (SCT) space. Color quantization is then achieved by center split algorithm applied on the color dimensions of the SCT space. Subsequently, Laplacian boundary detection operator is used to extract the contours of the ROI from the variegated coloring information. The system is implemented in Visual C++ and tested on temporal EMR color images of mouse. The system performs well giving perceptually reasonable segmentation of tumor, kidney and bladder of the mouse image. Experimental results with extensive set of EMR color images demonstrate the efficacy of the system developed.