On Depth Recovery from Gradient Vector Fields

摘要: Depth recovery from gradient vector fields is required when reconstructing a surface (in three-dimensional space) its gradients. Such reconstruction task results, for example, techniques in computer vision aiming at calculating normals (such as shape shading, photometric stereo, texture, contours and so on). Surprisingly, discrete integration has not been studied very intensively far. This chapter presents three classes of methods solving problems depth fields: two-scan method, Fourier-transform based wavelet-transform method. These extend previously known techniques, related proofs are given short but concise form. The method consists two different scans through field. final height values can be determined by averaging these scans. noniterative that boundary conditions needed, their robustness to noisy estimates improved choosing associated weighting parameters. overcomes the disadvantage which implicitly require function periodic. Experimental results using synthetic real images also presented.