Labeling algorithm is an image processing that marks individual connected regions to distinguish or separate. That is fundamental and important, especially in visiual recognition. One of them is the raster scanned labeling. This algorithm, however, takes a long time to mark respective regions consecutive numbers, as, for example, in the case of spreading branches or bending intricately. In this paper, first of all, we have investigated the problem, and then proved that a number of labels, called the evanescent label, were made for such regions and it was necessary to unify them from their connections. So we found the idea that we could decrese them by surrounding with the contour lines. And then this paper proposes the labeling algorithm using contour lines. Finally, simulation results comparison with the raster scanned algorithm are described.
輪郭線情報を利用したラベリング法
.png)
nagano_nct-26-03.pdf
(432.05KB)
