摘要 : Shape from focus is an elegant method that estimates the structure of a 3D object from a video of captured frames using the degree of focus as the principal cue. However, the quality of the estimated structure is vulnerable to sce... 展开 Shape from focus is an elegant method that estimates the structure of a 3D object from a video of captured frames using the degree of focus as the principal cue. However, the quality of the estimated structure is vulnerable to scene texture. The effect is particularly pronounced for objects that are smooth relative to the magnification of the optical system. In this paper, the shape estimation process is cast as an inverse problem. We exploit spatial dependencies by modeling the shape of the object with a discontinuity-adaptive Markov random field wherein the focus measure profile is used to judiciously control the degree of smoothness. The 3D information is obtained by minimizing a suitably derived energy function that preserves fine details of the underlying structure. We show by experimentation on several real-world specimens that our method yields state-of-the-art performance. 收起