* Feature tracking between image pairs using template matching.
* Full processing line from feature tracking to georectification.
* Distorted camera model. This allows the use of cheaper camera setups.
* No need for traditional image registration as a pre-processing step which can degrade the images due to resampling.
* Projecting between pixel and real world coordinates. 2D ↔ 3D
* Easily scriptable as it is a toolbox.
* Minimal dependencies.
* Inputs to feature tracking: two images
* Inputs to georectification: a DEM and some ground control points.
Very nice work removing MATLAB Toolbox dependencies.
Echoing @Chad's comments, the code provided here is clearly written and easily followed.
Thanks for sharing Aslak :-)
I have used the templatematch function extensively in my research. It efficiently performs feature tracking in a user-friendly manner. Each function in this package has example files which are easy to follow. ImGRAFT has saved me a tremendous amount of time and I have learned quite a bit by exploring the inner workings of templatematch. Thank you for writing and sharing these scripts, Aslak.
@mark: Thank you. I have been sweeping the code for dependencies today, and believe they are all gone.
Functions call padarray and fspecial which require the image processing toolbox. Unable to run demos without it.
@matthew: We have strived to make the toolbox free of dependencies. There may still be some dependencies in the demos, but there really should not be any in the toolbox itself.
If you stumble across some then please let us know where exactly, and we will try to remove them.
MATLAB toolbox dependencies? Mapping toolbox + ...
added an image
Inspired by: LMFnlsq - Solution of nonlinear least squares
Inspired: MODIS Mosaic of Antarctica