×

Study on a new adaptive two-dimensional Otsu image segmentation algorithm. (Chinese. English summary) Zbl 1240.68480

Summary: The threshold recognition functions of existing two-dimensional Otsu image segmentation algorithms are based on the trace count of between-cluster scattered measure matrices but they do not consider the cohesiveness of foreground and background pixels themselves and are too complex. Therefore, a novel algorithm with a new threshold recognition function is proposed. The algorithm counts the absolute differences regarding the class of object and background of an object image and adds them up to obtain the sum of total within-cluster absolute difference. Finally the algorithm sets the quotient of the sum of total within-cluster absolute difference and the total deviation as the threshold recognition function. Experimental results show the efficiency in segmentation and low cost in computation of the proposed algorithm with a newly defined threshold recognition function by subjectively and objectively comparing to the existing threshold recognition functions.

MSC:

68U10 Computing methodologies for image processing
94A08 Image processing (compression, reconstruction, etc.) in information and communication theory
PDFBibTeX XMLCite