Camera Calibration

Chris Broaddus
Imaging, Robotics, and Intelligent Systems Laboratory
The University of Tennessee

[Motivation] [Research Objectives] [Technical Approach] [Results] [Software] [Publications]

Motivation:

The motivation behind this research is to develop a robust geometric calibration process to calibrate an arbitrary camera. The model for radial and tangential distortion should automatically be selected which is the least complex, but sufficiently models the camera. 

Objectives:

The objective is to develop a statistical method for automatic model selection, including radial and tangential distortion, of an arbitrary camera.

Technical Approach:

We are extending Zhang's camera calibration algorithm to include tangential distortion. To determine the best model for radial and tangential distortion we are applying statistical methods, such as cross-validation, to automatically choose the best model for distortion.

Results:

The first image is the distorted image from an inexpensive camera and verifocal lens and the second image is after correction.

 Videos of distorted and undistorted (distorted, undistorted) 

Software:

These programs are not well documented! I can't guarantee them, but if you need help contact me (cbroaddu@utk.edu)

• SFM toolbox implemented in Matlab (incomplete) (download)

  
  Embedded in here under the m file "Script_Zhang" is an implementation of Zhengyou Zhang's camera calibration algorithm.

  
  The SFM toolbox, though incomplete, consists of Zhang's camera calibration technique. The difference being we added tangential distortion to the model and we do not limit the number of coefficients to model radial and tangential distortion. There are many other function in this toolbox for 3D reconstruction. This toolbox is being updated so it may not contain too much 3D stuff right now.

   

• Undistort in C code (download)
  
  Uses the calibration parameters from the calibration program above to undistort images fast.

Publications:

We are working on automatic model selection given an arbitrary camera using statistical methods. No publications available at this time.

This research is being conducted at the IRIS Lab by Chris Broaddus  under the supervision of Dr. Mongi A. Abidi.

Last updated:
Webmaster