This paper proposes an improved shape-from-focus (SFF) algorithm based on wavelet transform to realize rapid 3D surface reconstruction of laser damage via online microscopic imaging. Here, the depth was retrieved point by point through the 3D focus evaluation function, thereby reconstructing the 3D surface of the laser damage of optical elements. A 3D surface reconstruction experiment was performed on the damaged area of the silver mirror surface using the online microscopic imaging device for laser damage, based on the proposed algorithm. The results show that the definition ratio, sensitivity, and flat fluctuation of the proposed improved algorithm are 2.608, 0.131, and 0.356, respectively, and the relative error is 1.56% for the identified area with a maximum depth of 169.8 μm. As the proposed algorithm's results are compared to those of the traditional method, it is discovered that the definition ratio is increased by approximately one time, the sensitivity is increased by approximately three times, the relative error of depth measurement is reduced to one-tenth, and the flat fluctuation can be 1.1 times that of the conventional method. The proposed algorithm has already been employed in the development of an online microscopic damage detection system to achieve on-line rapid reconstruction and measurement for 3D laser damage topography.