The invention discloses a multi-threshold-value segmentation method based on a gray level histogram. According to the multi-threshold-value segmentation method, the maximum between-cluster variance principle is used, a dichotomy serves as the basis, and a minimal-value point in the gray level histogram serves as a reference point of multi-threshold-value segmentation. The multi-threshold-value segmentation method includes the steps of firstly, finding out a first threshold value of the gray level histogram according to the otsu method, secondly, respectively solving to obtain the maximum between-cluster variances of the two parts of the segmented histogram based on the dichotomy, thirdly, comparing the two variances, and finding out a threshold value corresponding the maximum variance to serve as a threshold-value point of next-time segmentation, fourthly, executing the three steps repeatedly until the given number of the threshold values is achieved, fifthly, finding out all wave trough points according to the smoothed histogram, and finally, comparing all the obtained threshold values with the wave trough points, and finding out the wave trough point closest to the threshold values to serve as the final threshold value. Compared with a traditional multi-threshold-value segmentation method, the multi-threshold-value segmentation method has the advantages that a part of segmentation noise is eliminated, and the segmentation effect and the adaptability are better than existing algorithms.