Welcome to Shufei's Homepage


Shu Fei Fan

Ph.D (2004 - 2010)
Center for Intelligent Machines
McGill University
Phone: (514) 398 - 1282
Email: fansf at cim dot mcgill dot ca

Right now, I am using vision methods to solve problems in the traffic industry.
I obtained my M.Eng. and B.Eng. (in Electrical Engineering) from Xidian University in 1998 and 1995 respectively. I used AI as a tool (e.g. Expert Systems) for 'data fusion' tasks involving signal processing systems. Before my studies at McGill, I worked as a software engineer and a system engineer between 1998 and 2003.

What's New:

July, 2010: I defended my PhD thesis (Wide-baseline stereo for three-dimensional urban scenes).
May, 2010: I started a full-time industry position.
March, 2010: I was awarded NSERC Industrial R&D Fellowship (IRDF) by the government of Canada.


My research interests fall under the areas of computer vision, image processing, and pattern recognition. My supervising committee members are Professor Frank Ferrie (supervisor), Kaleem Siddiqi and James Clark. I worked on such topics as feature extraction and matching, stereo, structure from motion, 3-D reconstruction, and camera calibration.

My PhD research is centered on the theme of extracting 3-D information from 2-D visual images, i.e. 'what can we say about 3-D structure of a scene from its images taken from different viewpoints'. One family is 'volumetric reconstruction' using calibrated images. These methods reconstructs a 3-D volume that is believed to be occupied by the 3-D scene, it uses such constraints as silhouettes, or photo-consistency. I proposed to fuse photo-consistency with stereo vision [1] to generate a more accurate model of the scene. Another approach, relying on structure from motion (SfM) algorithms, reconstructs 3-D models out of uncalibrated images. It is less constrained as one can collect the images with a hand-held camera while walking around the scene. Some of my works are concerned with establishing a geometric foundation for the SfM: how to extract distinct feature points [4] from the images and how to effectively match these feature points [2] across different perspectives, so that one can better obtain the multiple-view geometry [3].


Refereed Journals:

[1]. Fan, S. and Ferrie, F., Photo Hull regularized stereo, Image and Vision Computing, Volume 28, Issue 4, April 2010, Pages 724-730 (online first version)

Refereed Conference Proceedings:

[2]. Fan, S. and Ferrie, F., Context-Consistent Stereo Matching, ICCV workshop on 3-D Digital Imaging and Modeling, Oct. 2009. (PDF)
[3]. Fan, S. and Brooks, R. and Ferrie, F., Better Correspondence by Registration, ACCV, Sept. 2009. (PDF)
[4]. Fan, S. and Ferrie, F., Structure Guided Salient Region Detector, BMVC, Sept. 2008. (PDF)
[5]. Fan, S. and Ferrie, F., Photo Hull Regularized Stereo, Computer and Robot Vision, Jun. 2006. (Oral, PDF)


Fan, S. and Ferrie, F., Context-Consistent Stereo Matching, International Conference on Vision in 3D Environments, Toronto, June 2009
Fan, S. and Ferrie, F., Structure Guided Salient Region Detector, CIM REPARTI-Perception Seminar, Montreal, May 2008
Fan, S. and Ferrie, F., Image Modeling and Matching, Colloque REPARTI Workshop, Montreal, May 2008
Ferrie, F. and Lala, P. and Fan, S., ARTIFICIAL PERCEPTION LAB, CIM Lab demos at Colloque REPARTI Workshop, Montreal, May 2008

Teaching Experience

Teaching Assistent for the following courses at McGill University:

Fall, 2007: Topics in Vision and Robotics (ECSE 683). (graduate level, lab work design, tutor and project mentor)
Fall, 2005: Probability and Random Signals (ECSE 304).
Winter, 2005: Introduction to Computer Engineering (ECSE 221).

Training on teach skills: Learning to Teach Workshop, McGill University. Mar. 2009

Professional Service

Referee for international conferences:
ICCV 2005, ICPR 2005, ECCV 2006, ICPR 2006, ICPR 2008, CRV 2009, 3DIM 2009.

Referee for international periodicals:
Signal Processing Letter, March 2008.

IEEE student member
member of British Machine Vision Association (BMVA).

Courses taken

ECSE 626 Statistical Computer Vision (Winter 2004)
COMP 558 Fundamentals of Computer Vision (Winter 2004)
COMP 766 Shape Analysis in Computer Vision (Fall 2004)

In addition, I audited the following Computer Science courses: Pattern Recognition, Computational Geometry, Machine Learning, Algorithms and Data Structure.

Mailing address

Centre for Intelligent Machines, McGill University
3480 University Street, Room 410
Montreal, QC
H3A 2A7 Canada

Useful Links

Computer Vision Bibliography and more informaion maintained by Keith Price
The Middlebury Computer Vision Pages by Daniel Scharstein, Richard Szeliski and others
The performance evaluation of region detectors/descriptors by Mikolajczyk et. al.