3-D shape from contour and selective confirmation

Abstract

Surface contours are known to convey shape information about an object. The recovery of 3-D shapes from surface contours has been studied by many researches from theoretical points of view. Mathematically speaking, the information available in a single image is always ambiguous. Under certain regularity assumptions, local evidence about the shape of an object can often be derived. However, this local evidence, though rich in amount, is often inconsistent and not precise. A process called “selective confirmation” is proposed to resolve the ambiguities among the local evidence. Properties and extraction procedures of four types of contour are discussed in some detail. Experimental results of contour extraction and region grouping by selective confirmation are also presented.