Range image segmentation based on function approximation

Abstract

This paper considers the segmentation of range image measurements into surface patches which are either plane or curved and which are described formally by a function. After a formal description of the segmentation, we present and compare three methods suited for plane and curved patch segmentation and show the results of experiments conducted for testing their practical behaviour. The two first methods use the classical approach of region growing whereas the third method is based on a relaxation process. This original and last method exhibits simplicity and low computational complexity. Thanks to its parallel nature, it can be considered as a good candidate for range image segmentation in real-time applications. In machine vision, there is a growing interest in range imaging. This is mainly because, in contrast to the traditional intensity images, range images give the true geometric shape of the object surface, which is a very intrinsic feature of the object. Also, as a consequence of this interest, many range image sensors are already available, and more performant sensors under development. Range imaging is successfull where geometric but simple description of scene is needed and data interpretation straightforward, as for example in dimension control. However, in the case of more complex tasks requiring higher level interpretation of the scene, several processing steps are required: significant among those steps is the segmentation of range data into surface patches. Image segmentation is a well known problem in computer vision 1 . In the case of range images, segmentation is considered as the basic step by which range data of a scene is divided into several regions. Each region stands for a surface patch which is uniform with respect to a given property and which distinguishes it from the neighboring patches. Segmentation based on function approximation creates patches whose points are approximated by one function per patch. An interesting feature of this method is that the segmented image is a complete range image description from which the image data can be reconstructed.