Segmentation of 3-D range images of small industrial-like objects using the histogram of the surface mean curvature
Nabih N. Abdelmalek · International Journal of Systems Science · 1991
A procedure for the segmentation of 3-D range images of small industrial-like objects is presented. It is based on the idea that at segmentation boundaries, the surface mean curvatures are much higher than the mean curvatures of at least one of the regions which these boundaries separate. The segmentation boundaries arc such as folds or creases, ridge lines and the boundaries between planar and non-planar regions. The jump edges and the object(s) boundaries are obtained separately. The mean curvature at each surface point is calculated and the mean curvature histogram is obtained for the whole image, excluding the jump boundaries. From this histogram a coarse segmentation of the image is obtained. The segmentation boundaries may then be modified by examining the residuals of the fitting planes at each boundary point and its surrounding. Experimental results arc presented with comments.