Fast capsule image segmentation based on linear region growing

In computer vision based on-line capsule inspection, as defects contrasting with capsule are different in each part, it is needed to perform detection respectively in each of them. Yet owning to differential light penetration, the gray scale in the same region is inhomogeneous. Though various methods targeted at inhomogeneous region segmentation have been proposed in the literature, most of them are too complicated and time consuming to satisfy processing speed of on-line inspection. By analyzing the traditional image segmentation methods, a region growing technique based on linear scanning aiming at this issue is adopted. Unlike traditional region growing, it needs not to conduct complicated operation to specify initial seeds. The known input image structure is fundamental to its validity. Thus it is especially applicable to those applications which image structures have been already known, such as industrial on-line detection. It provides a fast extraction of the region-of-interest (ROI), saving precious time for following treatments. Its efficiency and robustness have been proved both in experiments and field applications.