Fast Multilevel Thresholding Based on Simulated Water Flooding

A fast multilevel thresholding method based on simulated water flooding in histogram is proposed. All possible thresholds (local minimums) are first detected and then sorted in descending order by their dynamics, which are calculated based on a simulated water flooding process in histogram. For N-level thresholding, the first N-1 thresholds of the sorted sequence are chosen. The optimal number of thresholds is achieved by doing a compromise between the discrepancy between the original and thresholded images and information compress. This method has some advantages: as the number of thresholds increases, the computation complexity keeps constant; when several results of different class numbers are required, only one time threshold searching is needed; automatically giving optimal number of thresholds will not greatly increase computation complexity. Experimental results on four testing images have proved its merits especially its high efficiency.