Seminar

Engineering HDR Seminar Series 12, 2011: Connected-Component Stereo Aggregation

Mr Ng Oon-Ee, PhD student from RAM


Date: 2011-09-27
Time: 11:00 to 12:00
Venue: Classroom 9-3-03


Abstract

Stereo vision algorithms can be divided into several general stages. Detailed analysis on each specific stage allows focused improvement of the entire algorithm. Existing stereo vision algorithms typically prioritize pixel and sub-pixel accuracy, focusing on the optimization stage, while commercial algorithms which prioritize real-time operation typically focus on the cost calculation stage. This paper presents a novel approach to cost aggregation, a module within the cost calculation stage, based on connected-component analysis. The conventional connected-component analysis algorithm is extended for use in the stereo vision matching cost aggregation problem, resulting in an algorithm which is comparable in efficiency to the best current aggregation algorithms while performing well at object boundaries, typically the most difficult problem for aggregation-based approaches.

About the Speaker

Ng Oon-Ee is currently completing his PhD candidature at Monash University Sunway Campus in Malaysia. He graduated from Monash University Malaysia with a Bachelor of Engineering (Mechatronics) degree with First Class Honours in 2006. He has been a student member of IEEE for four years. His research interests center primarily around stereo vision, with particular emphasis on algorithm development. He is also interested in computer algorithms in general and how they are applied to theoretical and practical problems.