Generalized Assorted Camera Arrays: Robust Cross-channel Registration and Applications

Jason Holloway, Sanjeev J. Koppal, Kaushik Mitra, and Ashok Veeraraghavan

Abstract: In this paper, we advocate for Generalized Assorted Camera (GAC) arrays for multi-modal imaging—i.e., a camera array with filters of different characteristics placed in front of each camera aperture. GAC provides us with three distinct advantages over GAP: ease of implementation, flexible application dependent imaging since filters are external and can be changed and depth information that can be used for enabling novel applications (e.g. post-capture refocusing). The primary challenge in GAC arrays is that since the different modalities are obtained from different viewpoints, there is a need for accurate and efficient cross-channel registration. Traditional approaches such as SSD, SAD, and mutual information all result in multi-modal registration errors. Here, we propose a robust cross-channel matching cost function, based on aligning normalized gradients, that allows us to compute cross-channel sub-pixel correspondences for scenes exhibiting non-trivial geometry. We highlight the promise of GAC arrays with our cross-channel normalized gradient cost for several applications such as low light imaging, post-capture refocusing, skin perfusion imaging using RGB+NIR and hyperspectral imaging.

Paper Links:

Download the Lo Res version (5.1 MB) or the Hi Res version (15.2 MB) of the paper.
Or find the paper on IEEE Xplore.

Citation:

Holloway, J.; Mitra, K.; Koppal, S.J.; Veeraraghavan, A.N., "Generalized Assorted Camera Arrays: Robust Cross-Channel Registration and Applications," Image Processing, IEEE Transactions on , vol.24, no.3, pp.823,835, March 2015

[+] BibTex Citation

@article{holloway2015generalized,
author={Holloway, Jason and Mitra, Kaushik and Koppal, Sanjeev J. and Veeraraghavan, Ashok N.},
journal={Image Processing, IEEE Transactions on},
title={Generalized Assorted Camera Arrays: Robust Cross-Channel Registration and Applications},
year={2015},
month={March},
volume={24},
number={3},
pages={823-835},
doi={10.1109/TIP.2014.2383315},
ISSN={1057-7149},
}

Code:

Download C++ source on Github

Select images from the paper:

Hover over images to see either the constituent images or the NIR enhanced images. Download the data on Github.

RGBY 1

RGBY 2

RGBY 3

NIR 1

NIR 2

Slides

Download the presenation accompaning this project as a PowerPoint slideshow [ppsx] (34 MB) or PDF (6 MB).

Poster

Download the poster (48" x 36") for this project: PDF (2.3 MB)