Unstructured Synthetic Aperture Photograph Based Occluded Object Imaging

Abstract

Recently the camera array synthetic aperture imaging has been proved to be a powerful technology for occluded object imaging. However, synthetic aperture imaging with a camera mounted on a moving platform is still a significant challenging task for many computer vision applications. This paper presents a novel unstructured synthetic aperture imaging approach to solve the above problem. The main characteristics of our approach include: (1) To the best of our knowledge, our proposed method is the first one for synthetic aperture imaging through occlusion on a moving platform. (2) In contrast to averaging data, we creatively use the clustering method to estimate the pixel color with high visible probability in the virtual focal plane. A new moving synthetic aperture imaging system has been built to capture unstructured light field in complex outdoor scene, and experimental results demonstrate that our approach successfully generates clear image of hidden object even under severe occlusion, which outperform the traditional synthetic aperture imaging method.