Grid reconstruction and skew angle estimation in Integral Images produced using circular microlenses

Nowadays a number of different three-dimensional (3D) systems compete in the field of capturing and delivering autostereoscopic (ASt) 3D content. Integral Imaging (InI) is a promising ASt technique that provides both horizontal and vertical parallax as well as high quality realistic 3D content. In this work we propose an InI preprocessing method for identification and accurate segmentation of the grid structure in Integral Images (InIms) generated using lens arrays (LAs) containing circular lenses. In the proposed method we utilize the gradient augmented circular hough transform to accurately detect circular regions in the acquired integral image (InIm). Subsequently by using a triangulation scheme followed by a statistical approach we accurately estimate the grid line structure of the utilized LA. This results in the accurate segmentation of the circular shaped elemental images (EIs) contained in the InIm, a process vital for the effectiveness of the InI methodology. We provide experimental results over artificial as well as optically acquired InIms to evaluate the accuracy of the method using objective metrics.