Recovering Depth from a Single Natural Image Based on Edge Blur Estimation

Recovering Depth from a Single Natural Image Based on Edge Blur Estimation

The traditional methods of Depth from Defocus (DFD) usually need to collect multiple defocus images, which are difficult to realize in practice. In this paper, the authors managed to solve this challenging problem with recovering the depth from a single image taken with an uncalibrated conventional camera. Different from all the existing depth recovering approaches, this approach avoides the collection of multiple images or the usage of deconvolution process, which provides a simple yet effective way to realize depth-recovering in a single image. The amount of defocus blur is obtained by the gradient magnitude ratio between the input and re-blurred images. Sparse blur map is obtained through the estimate of blur amount at the edge regions with the segmentation of images. Complete depth information is then recovered by propagating the sparse blur based on the local mean of edge blur. Experimental results on a variety of images show that the approaches in this paper can acquire a reliable estimation of the depth.