Perspective Correction of Distorted Projectors with an Uncalibrated Camera

Description

This project demonstrate image processing algorithms that allows the users to project extremely wide field distortioncorrected images on screens with horizontal discontinuity. Image pre-warping is used to achieve any non-parametric projector keystone correction and screen geometry correction. 

This project involves two casually placed projectors (Epson EX3240), a camera capturing the full projection field, two laptops and one flat or right-angled projection surface. A MATLAB graphical user interface (GUI) is created for user-friendly
application.

For wide field projection, two projectors are individually registered. A reasonable overlap of at least 5% of its total projection area is recommended for good linear blending result. This algorithm can be further developed for applications with projectors located at the back of the projection screen, multiple projectors with horizontal discontinuous surface, and simultaneous wide field video projection. With the development of curved displays, algorithm developed in this project can be used for large tiled high-resolution display.

Further work need to be conducted for color correction, pincushion and barrel distortion and other optical aberrations introduced by the hardware imperfection. This project has wide application in art visualization, face registration for make-up testing, virtual/augmented reality and Head-Up Displays.

References :

[1] A. Agarwal, C. Jawahar, and P. Narayanan. A survey of planar homography estimation techniques. Centre for Visual Information Technology, Tech. Rep. IIIT/ TR/2005/12, 2005.
[2] M. Brown, A. Majumder, and R. Yang. Camera-based calibration techniques for seamless multiprojector displays. Visualization and Computer Graphics, IEEE Transactions on, 11(2):193–206, 2005.
[3] A. Majumder and M. S. Brown. Practical multiprojector display design. AK Peters, Ltd., 2007. 

[4] R. Raskar and P. Beardsley. A self-correcting projector. In Computer Vision and Pattern Recognition, 2001. CVPR 2001. Proceedings of the 2001 IEEE Computer Society Conference on, volume 2, pages II–504. IEEE, 2001.
[5] P. Roman, M. Lazarov, and A. Majumder. A scalable distributed paradigm for multi-user interaction with tiled rear projection display walls. Visualization and Computer Graphics, IEEE Transactions on, 16(6):1623–1632, 2010.
[6] B. Sajadi, M. Lazarov, M. Gopi, and A. Majumder. Color seamlessness in multi-projector displays using constrained gamut morphing. Visualization and Computer Graphics, IEEE Transactions on, 15(6):1317– 1326, 2009.
[7] B. Sajadi and A. Majumder. Markerless viewindependent registration of multiple distorted projectors on extruded surfaces using an uncalibrated camera. Visualization and Computer Graphics, IEEE Transactions on, 15(6):1307–1316, 2009.
[8] B. Sajadi and A. Majumder. Scalable multi-view registration for multi-projector displays on vertically extruded surfaces. In Computer Graphics Forum, volume 29, pages 1063–1072.Wiley Online Library, 2010.
[9] A. Vedaldi and B. Fulkerson. VLFeat: An open and portable library of computer vision algorithms. http: //www.vlfeat.org/, 2008.

A variational multiphase level set approach to simultaneous segmentation and bias correction

XCode for Real Time Perspective Correction of Projector Image