Dense 3D Reconstruction from Severely Blurred Images using a Single Moving Camera

Motion blur frequently occurs in dense 3D reconstruc-tion using a single moving camera, and it degrades the qual-ity of the 3D reconstruction. To handle motion blur caused by rapid camera shakes, we propose a blur-aware depth re-construction method, which utilizes a pixel correspondence that is obtained by considering the effect of motion blur. Motion blur is dependent on 3D geometry, thus parameter-izing blurred appearance of images with scene depth given camera motion is possible and a depth map can be accu-rately estimated from the blur-considered pixel correspon-dence. The estimated depth is then converted into pixel-wise blur kernels, and non-uniformmotion blur is easily removed with low computational cost. The obtained blur kernel is depth-dependent, thus it effectively addresses scene-depth variation, which is a challenging problem in conventional non-uniform deblurring methods. 1