A digital radiographic image is segmented into various regions and a region thereof is further decomposed into subregions, wherein digital image data is acquired and subjected to multiple phases of digital imaging processes. The region decomposition method first uses progressively smoothing techniques to generate smoothed regions at multiple scales. Next, the number of connected components at each scale is computed for each smoothed region. A shape spectrum, which is the number of connected components as a function of scale, is subsequently constructed to determine the most stable range of scales and the most stable number of subregions into which the region is decomposed. Each pixel is then classified into subregions, according to the geometrical relationship of the connected components detected at most stable scales. Finally, a decomposed map is generated to function as multivalued templates for any further image processing to be done on various decomposed subregions.