Light Field Reconstruction Using a Planar Patch Model

Abstract

Abstract. Light fields are known for their potential in generating 3D reconstructions of a scene from novel viewpoints without need for a model of the scene. Reconstruction of novel views, however, often leads to ghosting artefacts, which can be relieved by correcting for the depth of objects within the scene using disparity compensation. Unfortunately, reconstructions from this disparity information suffer from a lack of information on the orientation and smoothness of the underlying surfaces. In this paper, we present a novel representation of the surfaces present in the scene using a planar patch approach. We then introduce a reconstruction algorithm designed to exploit this patch information to produce visually superior reconstructions at higher resolutions. Experimental results demonstrate the effectiveness of this reconstruction technique using high quality patch data when compared to traditional reconstruction methods.

Modulating the Size of Backprojection Surface Patches, in Volumetric Stereo, for Increasing Reconstruction Accuracy and Robustness

“…In [2] besides geometrical, photometrical properties are also computed on the patch and a multi-dimensional optimization is employed to determine the occupied voxels. In [4,5], spatially local maxima in the response of the operator are regarded as due to surface occurrence. It is argued that the proposed modulation can be directly adopted by volumetric methods, such as the above, that utilize a constant-size hypothetical patch.…”

“…In the second category, the size of the patch has not been modulated in the literature, except in [4,5] where a coarse to fine approach was employed in the evaluation of Equations 1 and 2, but for the acceleration of the computation. The modulation in these cases was, however, was discrete and identical for any location within the reconstruction volume and, also, irrelevant to the orientation of the patch.…”

“…In the second category [2,3,4,5], the acquired images are backprojected on a hypothetical planar surface patch, which is assumed at each voxel of the reconstruction volume. The backprojections are then compared as to their visual similarity, instead of regular patches in the acquired images.…”

“…[1]). More recently [2,3,4,5], stereo approaches have introduced a volumetric and more robust [6] way of performing this similarity-matching, by comparing the backprojections of the acquired images on hypothetical surfaces that are tangent to the imaged ones. This paper is related to the latter approach and its contribution consists in the improvement of the, above, volumetric similarity-matching approach in terms of accuracy and robustness.…”

Efficient, Precise, and Accurate Utilization of the Uniqueness Constraint in Multi-View Stereo

“…Its point correspondences can be, thus, located by searching for its most similar depictions in the acquired images. A number of works compensate for the projective distortion in the matching process (or else, match the textures in 3D), to obtain more matches and accuracy [7,8,9,10,11]. Treating the imaged surfaces as locally planar facilitates the backprojections of images at hypothetical planar patches and, in turn, the prediction that backprojections should match if the patch coincides with the surface.…”