Depth Estimation in Integral Imaging Based on a Maximum Voting Strategy

An approach that uses the scene information acquired by means of a three-dimensional (3D) synthetic aperture integral imaging system is presented. This method generates a depth map of the scene through a voting strategy. In particular, we consider the information given by each camera of the array for each pixel, and also the local information in the neighbourhood of that pixel. The proposed method obtains consistent results for any type of object surfaces as well as very sharp boundaries. In addition, we also contribute in this paper with a repository of a set of synthetic integral images generated by 3DS Max where the so-called ground truth (real-true depth map) is available. This resource can be used as a benchmark to test any Integral Imaging based range estimation method.