Local Descriptor for Dense Wide-Baseline Stereo Matching

	We show that it is possible to estimate depth from two wide baseline images using a dense descriptor. Our local descriptor, called DAISY, is very fast and efficient to compute. It depends on histograms of gradients like SIFT and GLOH but uses a Gaussian weighting and circularly symmetrical kernel. This gives us our speed and efficiency for dense computations. We compute 200-length descriptors for every pixel in an 800x600 image in less than 5 seconds.

We show that it is possible to estimate depth from two wide baseline images using a dense descriptor. Our local descriptor, called DAISY, is very fast and efficient to compute. It depends on histograms of gradients like SIFT and GLOH but uses a Gaussian weighting and circularly symmetrical kernel. This gives us our speed and efficiency for dense computations. We compute 200-length descriptors for every pixel in an 800x600 image in less than 5 seconds.

Results

Herz-Jesu Grid. By using two images, one from the left-most column and one from the upper row, we compute depth and occlusion maps from the view point of the row image. In the diagonal, we display the ground truth depth maps. We marked the correctly detected occlusions with green, incorrectly detected ones with blue and the missed ones with red. From this ﬁgure, it is apparent that DAISY can handle quite large baselines without losing too much from its accuracy.

Herz-Jesu Grid. By using two images, one from the left-most column and one from the upper row, we compute depth and occlusion maps from the view point of the row image. In the diagonal, we display the ground truth depth maps. We marked the correctly detected occlusions with green, incorrectly detected ones with blue and the missed ones with red. From this ﬁgure, it is apparent that DAISY can handle quite large baselines without losing too much from its accuracy.

Stereo reconstructions obtained using the DAISY descriptor. The first row are the input images where we use the first image and one of the other images as input and the second row shows the respective reconstructions where the first one is the laser scanned ground truth image.

Performance of DAISY against various transformations: Contrast+Rotation, Zoom, Blur and Viewpoint chage.

Download

Source code is available under the BSD License.

Matlab

USAGE-MATLAB
mdaisy-v1.0 DAISY has a matlab implementation now. // Wednesday, March 18, 2009 23:48:22 +0100

C++

USAGE | CHANGE-LOG | LICENCE

daisy-1.8.1.tar.gz minor bug fix //Sunday, October 11, 2009 11:42:22 +0200
daisy-1.8.0.tar.gz Licence change from GPL to BSD. //Tuesday, August 18, 2009 13:43:36 +0200
daisy-1.7.0.tar.gz bug fix + shared library support //Monday, July 13, 2009 14:41:00 +0200
daisy-1.6.1.tar.gz minor bug fix //Friday, July 10, 2009 10:45:08 +0200
daisy-1.6.tar.gz DAISY is faster now! // Wednesday, March 18, 2009 23:44:51 +0100
daisy-1.5.tar.gz IMPORTANT CHANGE. See Change.log //Thursday, February 26, 2009 18:21:43 +0100
daisy-1.4.tar.gz //Tuesday, February 17, 2009 16:01:55 +0100
daisy-1.3.1.tar.gz //Tuesday, February 10, 2009 14:53:03 +0100
daisy-1.3.tar.gz //Sunday, November 23, 2008 19:10:43 -0800
daisy-1.2.tar.gz //Thursday, July 3, 2008 14:43:34 +0200
daisy-1.1.tar.gz //Wednesday, May 28, 2008 14:24:56 +0200
daisy-0.0.tar.gz //Sunday, April 20, 2008 21:28:14 +0200

Relevant Publications

a fast local descriptor for dense matching

engin tola, vincent lepetit, pascal fua
june 2008 - proceedings of computer vision and pattern recognition (CVPR), alaska, usa

pdf
bibtex
data
source code
slides(9.5MB) //2008-07-10-11:23:13 +0200

daisy: an efficient dense descriptor applied to wide baseline stereo [preprint]

engin tola, vincent lepetit, pascal fua
march 2009 - ieee trans. on pattern analysis and machine intelligence
this is a preprint and may change in format, structurally or in small ways during the print process

paper
bibtex

Contact

Please do not hesitate for (reasonable) questions and bug reports to:

Engin Tola: name.surname@epfl.ch (obviously replace name and surname...)