Abstract: Image edge recognition is an important part of image processing. A gradient segmentation algorithm based on the depth-first traversal of images is presented. In this method, the data structure of the pixel is defined firstly, the four pixels around an arbitrary point of an image are estimated and the gradient values of the pixels are acquired. If the pixel satisfies the feature of an edge, the edge perpendicular to the directions of gradient is processed by depth-first traversal, and the pixels are marked at the same time. If there is no pixel to satisfy the feature in the direction, it will withdraw, and then the depth-first traversal will be implemented from the direction next to the maximum gradient, marking the pixel as corner, until the traversal of the whole image is completed. The experimental results show that edges and corners of an image can be segmented clearly and easily identified.