Multi-Level Image Thresholding Based on Histogram Voting

Thresholding is an easy yet efficient method in image segmentation, when dividing different objects with distinct gray-levels. Its main problem is how effective the thresholds divide the image. A new multilevel thresholding method is proposed in this study, which bases on voting response of all histogram bins to each bin. Smoothed the histogram, the method accumulates all voting of other bins by a novel measure function which integrates many factors, such as the difference, the distance and the value of histogram bins. Then, final thresholds are obtained by the extremum and the percentage. The method prefers valleys, and it is efficient because of the variable step and the rule which stop the unnecessary voting process. In the experiment, the method can works well both in clear and noisy images, and its effectiveness is also demonstrated by some comparisons with other methods.