The invention discloses an image edge detection method based on window modified multi-level median filtering, which includes: 1, adding noise in an original image, and selecting a rectangular window; 2, on the basis of multi-level median filtering algorithm, dividing the rectangular window into four sub-windows which respectively indicate four one-dimensional windows in different directions, wherein window trend selection makes full use of correlation information of directions; 3, selecting size of a template, and respectively selecting median of pixels of corresponding windows for the four filtering sub-windows; 4, initially defining image output results after the windows are modified, and obtaining an initial edge outline of the edge of the image; and 4, selecting appropriate parameters and thresholds according to characteristics of the processed image, and generating a final target edge detection image. The image edge detection method based on the window modified multi-level median filtering not only well restrains the noise, but also effectively detects the edge of the image, shows good effects, and can be applied to detection of image processing, computer vision and industrial automation.