A fast mosaic method of surface defects based on sparse matrix is proposed. In the method, a ring white light source is used to irradiate uniformly on the component surface under test, and thus a bright defect image is formed in a dark background by a microscopic scattering dark field imaging system. Then the sub-aperture stitching image is obtained by scanning the x and y directions of optical elements. Based on sparse matrix and image stitching, the fast stitching of sub-aperture images is carried out to obtain a full-aperture defect image. Based on the principle of minimal external rectangle, the image defects are identified and classified. Seven defects on the surface of optical components are finally obtained, in which the largest one has a length of 15.2110 mm and a width of 0.0297 mm. Simultaneously there exist five pock marks, in which the largest one has a length of 0.1089 mm and a width of 0.0967 mm. The relative error range of scratch width is -5.00%-5.50% after comparison between the measured and standard scratch widths. On this basis, the actual optical surface is detected, and the surface defect information of optical elements is obtained.