The invention discloses an object contour extraction method based on sparse representation, which belongs to the technical field of digital image processing and perception. The method comprises the following steps: a) determining an initial contour for an inputted object image in a training set; b) carrying out multiple combined treatment on the initial contour, wherein each combined treatment comprises local search and correction; and c) terminating the combined treatment and outputting the obtained contour when the times of the local search achieves the set threshold or the number of control points which do not move the positions during one local search exceeds the set threshold. The method can be widely applied in various fields of image understanding and perception, improve the efficiency of automatic image and video processing of a computer and have broad application prospects.