Fast Image Segmentation Method Based On Threshold

Tang Xu-Dong,Pang Yong-Jie,Zhang He,Zhu Wei

DOI: https://doi.org/10.1109/CCDC.2008.4597936

2008-01-01

Abstract:Image segmentation is a key part in image processing field Though traditional threshold methods based on maximum entropy principle could sometimes divide the image into object and background, the size of search space increase very rapidly when the number of parameters needed to determine the membership function increase, so it's time-consuming is often an obstacle. In this paper, based on the traditional threshold methods, the membership function is simplified and a new recursive scheme is introduced which decreases the computation complexity of the traditional method. The approach does not need the calculation of the membership function and increases the calculated speed. The experiments prove this novel approach is not only segment accuracy, but also decrease the computation time. Comparing with the traditional methods, the new method shows better real timing and noise restraining performance.