Guided Filtering based Pyramidal Stereo Matching for Unrectified Images

Xu, T. and Cockshott, W. (2015) Guided Filtering based Pyramidal Stereo Matching for Unrectified Images. In: IVCNZ 2015  Image and Vision Computing, Auckland, New Zealand, 23-24 Nov 2015,

112151.pdf - Accepted Version



Stereo matching deals with recovering quantitative depth information from a set of input images, based on the visual disparity between corresponding points. Generally most of the algorithms assume that the processed images are rectified. As robotics becomes popular, conducting stereo matching in the context of cloth manipulation, such as obtaining the disparity map of the garments from the two cameras of the cloth folding robot, is useful and challenging. This is resulted from the fact of the high efficiency, accuracy and low memory requirement under the usage of high resolution images in order to capture the details (e.g. cloth wrinkles) for the given application (e.g. cloth folding). Meanwhile, the images can be unrectified. Therefore, we propose to adapt guided filtering algorithm into the pyramidical stereo matching framework that works directly for unrectified images. To evaluate the proposed unrectified stereo matching in terms of accuracy, we present three datasets that are suited to especially the characteristics of the task of cloth manipulations. By com- paring the proposed algorithm with two baseline algorithms on those three datasets, we demonstrate that our proposed approach is accurate, efficient and requires low memory. This also shows that rather than relying on image rectification, directly applying stereo matching through the unrectified images can be also quite effective and meanwhile efficient

Item Type:Conference Proceedings
Glasgow Author(s) Enlighten ID:Xu, Professor Tianrui and Cockshott, Dr William
Authors: Xu, T., and Cockshott, W.
Subjects:Q Science > QA Mathematics > QA76 Computer software
College/School:College of Medical Veterinary and Life Sciences > School of Molecular Biosciences
College of Science and Engineering > School of Computing Science
Copyright Holders:Copyright © 2015 The Authors
Publisher Policy:Reproduced in accordance with the copyright policy of the publisher
Related URLs:

University Staff: Request a correction | Enlighten Editors: Update this record