A region of interest labeling algorithm using three mask patterns

Abstract

Labeling is one of the most basic and important processes in image analysis, which is used to identify detached objects by assigning the same number (labels) to all adjacent connected pixels in a binary image. Labeling algorithms have long been studied, and a variety of algorithms have been developed. Two scans method is easy to implement hardware. The two scans method requires memory for 1-D and 2-D tables to perform labeling. In this paper, three masks are used to assign label values to minimize memory usage, and an algorithm to increase computation speed by separating the inputted image into regions of interest and non-interest is proposed. As a result of experiment that is continuous image of 100 frames, Assigned provisional label is that conventional algorithm is 7657, [9] is 14665 and proposed algorithm is 5710. Processing times is required of conventional algorithm 341.6 ms, [9] 621.328 ms, proposed algorithm 275.18 ms. To verify the performance of the proposed algorithm, an experiment has been performed using a variety of binary images. © 2013 Springer Science+Business Media Dordrecht.