The valley-emphasis method and the neighborhood valley-emphasis method all consider the gray threshold, where the valley and the neighborhood values affect the probability of segmentation results. This paper proposes an improved Otsu algorithm, which combines valley gray value and spatial distribution of the image, and constructs a new standard deviation factor, thus forms a new threshold segmentation of the target formula. Through the simulation experiment, the proposed algorithm can effectively segment the images, which has a very strong practicability.