关键词: 图像分割, 阈值法, 交叉熵, 卡方散度, 关键阈值

Abstract: Traditional cross-entropy thresholding method is sensitive to noise and with much longer computation time. In order to improve the performance of this algorithm, a novel image thresholding segmentation criteria based on 2-D minimum chi-square divergence is proposed. A novel 2-D histogram based on improved median filter is structured. The difference between the segmented image and the original one is measured by the symmetric chi-square divergence. The neighborhood image is segmented with the key threshold to obtain better segmentation effects. Experimental results demonstrate that the proposed method’s computing time is much less, and its segmentation effect and anti-noise are better than 2-D minimum cross entropy and 2-D Otsu.

Key words: image segmentation, thresholding method, cross entropy, chi-square divergence, key threshold