Method and Processor for 3D Scene Representation

摘要

The present invention relates to three dimension (3D) scene representations and in particular to a method and a processor for providing improved 3D scene representations. An objective of the embodiments of the present invention is to improve the determination of consistency among a plurality of projections at a virtual view denoted νF. When determining the consistency, entries of a distance matrix, indicative of distance differences of 3D components between different views for a corresponding segment k when projected to the predefined view (νF) for each segment k, are compared with entries of a threshold matrix. The objective is achieved by assigning each segment k of a 3D component to a cluster based on individual rules for each cluster and one threshold matrix and by determining one threshold matrix for each cluster based on the segments of that cluster.

主权项

1. A method in a processor for representing a 3D scene, comprising:
projecting to a predefined view a 3D component of the 3D scene to be represented, captured at least three different views, determining a value for each of the projected 3D component captured at least three different views, defining clusters, wherein each cluster will be assigned a set of segments of a 3D component, defining for each cluster an individual rule for assigning segments of a 3D component to the respective cluster, assigning each segment of the 3D component to a cluster based on the individual rules, detecting consistency among the projected 3D components captured at least three different views by
comparing entries of a distance matrix, indicative of distance differences between the at least three different views for the corresponding segment k when projected to the predefined view for each segment k, with a threshold matrix, wherein the threshold matrix is determined based on the segments of the cluster, anddetermining consistency values for each cluster individually, using the determined values and the consistency value when representing the 3D scene at the predefined view.