Keypoint extractors refers to visual modules that are aimed at detecting salient and highly distinguisable points and/or regions in the image. This page provides a MATLAB/Mex implementation for the sRD-SIFT developed by Miguel Lourenço and João P. Barreto. The demos can be accessed in the form of a zip file containing the C/Matlab demo code. Our method is a model-based approach that requires that you roughly calibration your data set. Have a look at the Calibration section given below to know how to easily calibrate your images in a few seconds.
To use our code for benchmark comparision please request executables of our C version since they are more accurate and considerably faster.
If you use any of the code provided please cite our TRO paper.
We provide a set of comparative test images acquired by cameras with real lens distortion . The frames were acquried using a lens with low distortion (RD = 10%), a 4mm minilens commonly used for robotics' applications (RD = 25%), and a fish-eye lens with a wide field-of-view (RD = 45 %).
M. Lourenco, J.P. Barreto, and F. Vasconcelos, "sRD-SIFT Keypoint Detection and Matching in Images with Radial Distortion", " IEEE Transactions on Robotics 2012.
M. Lourenco, J.P. Barreto, and A. Malti, "Feature Detection and Matching in Images with Radial Distortion," IEEE International Conference on Robotics and Automation, 2010. (Best Student Paper Award Finalist)
The sRDSIFT is a model-based filtering approach to perform image keypoint detection. As so, to be able to run it you need to known the calibration parameter qsi and the distortion center.
The best way to compute the required calibration is the SIC algorithm of Barreto et al. for images with radial distortion