The Laplace integral for a watershed segmentation

Introduces a new powerful feature for the merging and classification of regions segmented by a watershed transform. It is based on the contour integral over watershed lines, which is equivalent to the area integral of Laplace values of the enclosed region. This approach is appealing since it complements the gradient-input image for the watershed segmentation with directional information and has low computational costs. The authors also show that the area integral of a Laplace image can be used to generate a new texture measure where undesirable contributions of region boundaries are eliminated. The approach is demonstrated in segmentation of radiographic images into shutter, direct radiation, and metal implant areas.