Granulometry based detection of junction and end points in patent drawings

In this paper, we propose an optimum detection of junction points and end points in patent drawings using morphological spurring and the granulometric curve of the image. The detection of these features in the skeletonized image gives rise to false detections due to skeletonization noise. An intersection of a spurred version of the skeleton image with the image containing junction points detection eradicates such false detections. To determine the optimum number of iterations for the spurring operation, the proposed approach takes into account the average thickness of lines found in the original image using the granulometric curve. The proposed method is parameter free, scale invariant and locates the positions of the junction points and end points accurately. We create ground truth of the junction points and end points and obtain the detection performance of the proposed method.