Apparatus and method of reducing noise in range images

Abstract

An apparatus and method of reducing noise in range images, the apparatus including: a maximum likelihood detection unit calculating an equation of observation points disposed in a plurality of grids that constitute range images representing three-dimensional (3D) coordinate information, to satisfy maximum log-likelihood on each coordinate axis in a coordinate space represented by the range images based on probability distribution of coordinate values assigned to each of the observation points and neighboring observation points; a linearization unit transforming equations calculated on each of the observation points into a linear equation that is a linear system on each coordinate axis in a coordinate space; a constraint detection unit calculating a constraint equation in which a unit tangent vector of each observation point and a unit tangent vector of a neighboring point are identical with each other on each coordinate axis of the range images; and a noise reduction unit reducing noise in range images by applying a resultant value that is obtained by calculating a value of a linear equation represented in the form of least squares together with the constraint equation by using a normal equation on a position of each observation point in the range images. Noise in range images in which noise is not uniformly distributed, may be well reduced so that a smooth curved surface having stable curvature may be obtained.