The proposed novel method categorizes candidate boundaries into visually-prominent and non-prominent boundaries, considering local intensity cues of multiple color channels and pixel-prominence-values measured as a function of proximity-influence of other contours. The results of the method with and without incorporating a wavelet transform are compared for images of different characteristics, types and resolutions. The method has been exhaustively tested on textured, medical, natural, biometric and synthesized images for prominent boundaries detection. The method results are qualitatively compared with those of human segmented images of benchmark image-segmentation dataset. The results presented show suitability and compatibility of the method for detecting prominent boundaries in various images. These boundaries - forming regions, make the basis for object detection and identification.