Synthetic aperture based on plenoptic cameras for seeing behind occlusion

Abstract

In this paper, a synthetic aperture imaging algorithm is proposed to see the object behind occlusion based on plenoptic cameras. Depth estimated from the plenoptic image is used to filter out feature points on occlusions. Two handshake principal is proposed to pair feature points from central sub-aperture image with those from other sub-aperture images. Homography transformations are derived among the sub-aperture images to project other sub-aperture images to the central one. By averaging the images after projection, a synthetic aperture image is generated to smear out the occlusions. Comparing with existing works, the proposed algorithm provides synthetic aperture image with higher sharpness and clearer de-occlusion effect.