B-rep from unregistered multiple range images

Abstract

We present a method to produce an integrated description of an object given multiple range views without registration. Our multiple view representation is in the form of B-rep (boundary representation), which has not been done so far by the computer vision community. We describe each view of the object as an attributed graph whose nodes are surface patches and links are the relations between surfaces. Any two attributed graphs, each corresponding to a given view, are matched and the rigid motion transformation between them is computed. The basic strategy for multiple view integration is composed of two aspects: We first build a composite graph, which contains the bounding surfaces, and their corresponding attributes, and then intersect these surfaces so that the edges and vertices corresponding to the B-rep description are identified. We present results on objects with polyhedral as well as quadratic curved surfaces.

Principal component analysis with missing data and its application to object modeling

“…Ferrie and Levine [9] merged multiple views using correspondence points which are identified by correlation over the differential properties of the surface. Parvin and Medioni [12] proposed to construct boundary representation (B-rep) object models from unregistered multiple range images. Each view of the object is represented as an adjacency graph where nodes represent surface patches with attributes and arcs representing adjacency between surfaces.…”

“…Given two adjacent segmented images I, and 12, we would like to find correspondence between different regions in two views, i.e., we want to find a mapping ý: (11 -- 12) such that a certain distance measurement d (I1, 12) is minimized.…”

“…Most previous systems have attempted to apply inter-frame motion estimates to successive pairs of views in a sequential manner [ 12]. Whenever a new view is introduced, it is matched with the previous view, and the transformation between these two successive views has to be recovered before the object model is updated.…”

Shape modeling from multiple view images using GAs

“…Especially, virtual world is signicant subject for communication using computer networks. Many previous modeling method use range data of the object shape which can generally be obtained by range nders [4] [5].However, range nders are not popular tool, and they can not obtain the surface texture data which can be used for synthesizing images presented in virtual space (image-based rendering).…”

Extracting surface patches from complete range descriptions

“…The problem of merging multiple range datasets has received considerable research [2,5,13,7,8,12,20,21,24,25,26,27,28,29,30] in the past few years, so we will not review it here. However, the result of this research is that it is now possible to obtain quite good registration of clean range image datasets of smooth curved and developable surfaced objects.…”