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.

Light fields can be used to generate photorealistic renderings of a scene from novel viewpoints without need for a model of the scene. Reconstruction of a novel view, however, often leads to ghosting artefacts, which can be relieved by correcting for the depth of objects within the scene using disparity compensation. Disparity estimation offers a solution to both better reconstruction and compression of large amounts of data in light fields. In this paper, we present two novel methods of disparity estimation for light fields: a global method based on the idea of photo-consistency and a local method which employs wavelet subbands for initial disparity estimation and Kalman filtering to refine the estimates. Experimental results demonstrate the effectiveness of the two methods as compared to other photo-consistency based disparity estimation techniques.

No abstract

No abstract

This paper describes a novel method to perform video based rendering. By capturing a set of real video sequences of a scene, the aim is to render a video sequence, in real time, from any viewpoint. By modelling the surfaces of a scene as a set of disjoint planar patches, we are able to efficiently estimate the parameters of the scene geometry. The patches can then be tracked over time using a multiresolution hierarchy. This time-varying surface model, and the images, are the input for the rendering algorithm, which uses a fuzzy z-buffer and projective texturing to generate reconstructions.