Anisotropic feature detector
----------------------------

--Syntax--

anisotropic.exe [-cpu/-gpu] [-i/-v] [filename] [sigma] [step] [threshold] [number of scales] [number of frames]

-cpu/-gpu: Flag indicating whether the processing is carried out by CPU or CUDA-compatible GPU.
-i/-v: Flag indicating whether the file is an image (-i) or a video (-v).
filename: Full path of file to be processed.
sigma: radius of kernel for the initial scale.
step: interval between successive scales, defined as \xi\ in section 2 of [1].
threshold: cornerness threshold as defined in equation (3) in [1].
number of scales: number of scales to be considered.
number of frames: If the -v flag is selected, number of frames to be processed in the video.


--Output format--
Output is stored in [cpu/gpu][Video/Img]Out_xxx, where xxx is the number of the frame being processed. The output file will contain information about the feature position (x,y), the ellipse parameters (A,B,C), the feature's cornerness value (cornerness), the radius of the ellipse along the major axis (r1), the radius of the ellipse along the minor axis (r2), the ellipse orientation.

--Example--

CPU image processing
anisotropic -cpu -i img1.bmp 1.5 1.44 0.3 7 1

GPU video processing
anisotropic -gpu -v video1.avi 1.5 1.44 0.3 7 1

--References--

When using the software, please cite the following papers:

[1]  S. Giannarou, M. Visentini-Scarzanella, G.-Z. Yang, Probabilistic Tracking of Affine-Invariant Anisotropic Regions, IEEE Transactions on Pattern Recognition and Machine Intelligence (TPAMI), 2012.
[2]  S. Giannarou, M. Visentini-Scarzanella, G.-Z. Yang, Affine-invariant anisotropic detector for soft tissue tracking in minimally invasive surgery, Proceeding of IEEE International Symposium on Biomedical Imaging (ISBI) 2009, Boston, MA.