The invention provides a new method of digital image prospect target extraction based on pixel heredity. The method comprises the following steps of (1) using an Ncut technology and a Gaussian variation model to extract a key prospect from an image so as to form an initial population; (2) using an edge detection method to extract and acquire edges of an original image and the key prospect, taking a common part, and using connectivity and a pixel similarity to determine a heredity boundary (a target outline); and (3) constructing an adaptability function, selecting pixels with high adaptability as parents, and through intersection, variation and other heredity process, regenerating the new prospect target in the image. The process of the whole method is full-automatic and any man-machine interaction is not needed. A lot of experiments prove that the method is very effective and an accurate experiment result can be acquired.