Recovering Depth Map from Enhanced Image Gradients

This paper presents a new and efficient approach for recovering surface height from its image gradients. In order to compensate noise effect in image intensities, Gaussian filter is used to smooth the intensity values. Image gradient is calculated from the smoothen image by taking the central difference of intensity values. A minimization problem is formulated in terms of gradient values. This minimization problem is solved in Fourier domain to avoid the iterative minimization. The experiments have been performed on real and synthetic images to demonstrate the accuracy of the approach using Gaussian filter. Several different measures have been used for error estimation.