Gradient guided feature selection in stereo matching

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Latest researches in stereo matching mainly focus on disparity optimization. Particularly, much effort has been put into the cost aggregation to improve the cost computation accuracy. However, the performance of the cost aggregation is determined by both the aggregation method and the prior cost volume obtained by the matching cost computation based on certain image features. Different choices of features could result in different quality cost volumes. Therefore, it is necessary to study feature selection for stereo matching. Our experiments show that in the region close to the edges or with high textures, features captured in smaller windows could lead to a better performance. Based on this observation, a novel gradient image guided feature selection method is developed, which takes the gradient image of the left frame as a reference image to regulate the image feature selection. Experiments based on Middlebury benchmark have verified the superiority of the proposed algorithm.

Original languageEnglish
Title of host publicationElectronics, Communications and Networks IV - Proceedings of the 4th International Conference on Electronics, Communications and Networks, CECNet2014
EditorsAmir Hussain, Mirjana Ivanovic
Pages1545-1548
Number of pages4
DOIs
StatePublished - 2015
Event4th International Conference on Electronics, Communications and Networks, CECNet2014 - Beijing, China
Duration: 12 Dec 201415 Dec 2014

Publication series

NameElectronics, Communications and Networks IV - Proceedings of the 4th International Conference on Electronics, Communications and Networks, CECNet2014
Volume2

Conference

Conference4th International Conference on Electronics, Communications and Networks, CECNet2014
Country/TerritoryChina
CityBeijing
Period12/12/1415/12/14

Fingerprint

Dive into the research topics of 'Gradient guided feature selection in stereo matching'. Together they form a unique fingerprint.

Cite this