An Overlapping NAM Segmentation Approach for Grayscale Images and Its Applications to Moment Calculation
Yunping Zheng, Tao Liu · 2023
Moments are widely used in image classification, pattern recognition and scene analysis because of their translation, rotation and scaling invariance, so it is of great theoretical importance and application to study moment calculation. According to the fast moment calculation formula derived from integration, the moment value of a binary image can be calculated in O(N) time as long as the coordinates of the upper left and lower right vertices of each sub-pattern after its division can be obtained, where N is the number of sub-patterns. In recent years, many excellent algorithms have emerged to transition from the study of moment computation of binary images to the study of moment computation of grey images, including the moment computation algorithm based on the image block representation (IBR) block scan and the moment computation algorithm based on non-symmetry and anti-packing model (NAM) block scan, etc. In essence, all these algorithms try to reduce the number of subpatterns to be divided as low as possible through the greedy algorithm, so as to reduce the time consuming moment computation. Since NAM is a raster scanning method and can approximate the optimal result and reduce the number of subpatterns significantly, and the overlapping NAM block scanning method (ORNAMC) can further reduce the number of subpatterns. In this paper, we proposed an ORNAMC-based fast moment calculation algorithm, which can theoretically and experimentally improve the efficiency of moment calculation.