Three-dimensional object delineation by dynamic programming
J.P. Pascual Starink, Jan J. Gerbrands · Bioimaging · 1994
State space search techniques like dynamic programming have been proposed for the delineation of object regions in extremely noisy two-dimensional images. Usually, such methods are applied directly to the image or to a window within the image. The method described here starts with a region of interest enclosing the required boundary. A geometric transform is used to straighten the region of interest into a rectangular matrix. In the transform domain, merit coefficients are computed and a dynamic programming optimal path algorithm is applied to find the transformed object region boundary. Finally, the path is transformed back to the original image domain. The subject of the current paper is the extension of the two-dimensional search method to three dimensions. Experimental results are given, as well as an example from the field of cell biology.