An efficient two-scan algorithm for computing basic shape features of objects in a binary image

Abstract

The basic shape features of an object in a binary image, i.e., the area, perimeter, circularity, and centroid, are important for image analysis and pattern recognition. In conventional algorithms, to calculate the basic shape features of objects in a binary image, it is usually necessary to first perform connected-component labeling to generate a labeled image (intermediate image), in which every image object is assigned a unique label so that it may be distinguished. Using the labeled image, the basic shape features of the object corresponding to each label can then be calculated. When a two-scan labeling algorithm is used, three scans are necessary. This paper proposes an efficient algorithm for calculating the shape features of objects in a binary image. Instead of a labeled image, our proposed algorithm calculates the basic shape features of objects using the image and the representative label table generated by the first scan of an efficient two-scan labeling algorithm. Thus, we can compute shape features using two scans. Experiments demonstrate that our proposed algorithm is much more efficient than conventional algorithms for calculating the basic shape features of objects in a binary image.