A novel cloud surface shape estimation method based on SFS
Jiayue Hou · 2016
Cloud is a common element of the sky and plays an important role in outside scene modeling. In this paper, a novel cloud surface shape estimation method from an image is proposed based on Shape from Shading(SFS). Shape from shading(SFS) algorithm is inherently ill-posed for the one image, which means that no shape or more shapes may synthetic the same input image. To estimate the cloud surface shape in the image, we propose two priors to make the problem well-posed: 1) the average depth prior, imposed on the average depths of different areas of the cloud; 2) the boundary prior, including the occluding contour and self-occlusions. Then, we formulate an optimization problem based on the shape from shading(SFS) algorithm. To solve the optimization problem, we utilize an improved L-BFGS method. The experimental results show that our method can estimate cloud surface shape similar to that of the image.