image_mosaic_stitching
image_mosaic_stitching copied to clipboard
Implemented a pipeline for 2D image mosaic and stitching. Feature extraction & matching, Adaptive Non-Maximal Suppression (ANMS), geometric blur, RANSAC
Image Mosaic and Stitching
-Yiren Lu (luyiren [at] seas [dot] upenn [dot] edu)
A matlab implementation of image mosaic and stitching:
Harris corner detector
see corner_detector_impl.m
Harris, Chris, and Mike Stephens. "A combined corner and edge detector." Alvey vision conference. Vol. 15. 1988.
Adaptive Non-Maximal Suppression (ANMS)
see anms.m
Brown, Matthew, Richard Szeliski, and Simon Winder. "Multi-image matching using multi-scale oriented patches." 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05). Vol. 1. IEEE, 2005.
Geometric blur
see geo_blur.m
Berg, Alexander C., and Jitendra Malik. "Geometric blur for template matching." Computer Vision and Pattern Recognition, 2001. CVPR 2001. Proceedings of the 2001 IEEE Computer Society Conference on. Vol. 1. IEEE, 2001.
Image descriptors matching
see feat_desc.m
or feat_desc_geoblur.m
- 40x40 patch descriptor downsampled to 8x8
- Normalized to mean of 0 and std of 1
- SSD of 1st matching / SSD of 2nd matching < 0.6
Outlier rejection by RANSAC
see ransac_est_homography.m
- Estimate homographic transformation matrix with 4 random feature points.
Image homographic warping and stitching
see mymosaic.m
- Pixel mapping by the homegraphic matrices
Image blending (distance to boundary)
see mymosaic.m
- Alpha = dist2border1/(dist2border1+dist2border2)
Run Code
- See
demo.m
- More datasets could be downloaded here: https://drive.google.com/file/d/0Bw-gLvstCiC-T1JfeGdqQ2JBOTA/view?usp=sharing
Presentation Slides
More Results
License
The contents of this repository are licensed under the MIT License.