Stereo Segmentation based on Adaptively Linked Hierarchical Structures

J.A. Rodríguez, P. Camacho, C. Urdiales, C. Matías, and F. Sandoval (Spain)

Keywords

hierarchical segmentation, adaptive stabilization, disparity, vergence

Abstract

Segmentation is a tool which has been widely used in image processing and computer vision for a variety of applications. The process consists in dividing a scene into a number of homogeneous regions with respect to one or several of its features. One of these characteristics is the depth at which objects are located. This information is implictly included in stereo pairs of images. Stereo matching and feature-based techniques work only under constrained conditions. Hierarchical structures, commonly named pyramids, have been typically used to reduce the computational load required for image processing. Adaptively linked pyramids have been successfully used in order to obtain compact regions as a result of the segmentation process for static images and video sequences. This paper presents a new stereo matching procedure to segment a stereo pair of images into homogeneous regions attending to the grey level and the disparity information contained in the stereo imagery.

Important Links:



Go Back


IASTED
Rotating Call For Paper Image