fast SUSAN feature detection and non-linear filter

a fast implementation of the SUSAN feature detector by non-linear filter

Description

This is a fast implementation of the SUSAN feature detection algorithm described at:

www.fmrib.ox.ac.uk/~steve/susan/

Also included is a general purpose non-linear filter based on the differences in intensities in a neighborhood. Any implementation of the SUSAN algroithm will require some kind of non-linear filter, so you may use this one if you don't have one already.

You will have to compile (this is what makes it fast) nlDiffFilter.f90 with mexf90.f90 to run everything else. I got some warnings from the compilaer about using the -fast option, but this hasn't been a problem.

You may try running demoSUSAN.m in debug mode to get a feel for how everything works.

• MATLAB release: MATLAB 7.0.1 (R14SP1)
• Other requirements: You will need a fortran90 compiler, although it could probably be modified forfortran 77 with some work. But if you were going to do this you might as well write your own version in C.

[Peter Staab]

Hello,
Can you please give me a short tutorial on how to compile the fortran file?
Thanks,

[Peter Staab]

In line 34 (tempA(1+r1:m+1+r1,1+r2:1+n+r2) = A ) in "nlDiffFilter.f90" an exception is raised when I run your demo code.

[rox ddd]

 

[tong qing-bin]

 

[lin ling]

HOW DO YOU DO!

[Anthony]

I'd really like to use this, however, as others have mentioned I can't find a way to compile the Fortran file. Any advice?

[Anthony]

 

[Yu Yanan]

What is the type of noiseI as the input image?

[suruchi goel]

I am unable to use this package. Please explain the compilation of .f90 file.
Do i have to install some compiler ..??
Waiting for your kind reply.
Thanks

[Madeena Sultana]

I am also unable to use this package. How to compile fortran files in MATLAB? Please help.

[Richa Sharma]

I am not able to compile the code in matlab.Error I get is:
Undefined function or method 'nlDiffFilter' for input arguments of type 'double'.

Error in ==> getUSAN at 38
U = nlDiffFilter(A,table,mask)/maxResponse;

Error in ==> SUSAN at 30
USAN = getUSAN(A,maxInt,sigma,thresh,width);

Error in ==> demoSUSAN at 6
[U,V] = SUSAN(noiseI,255,2,64,.55,2);

[Ahmed Farrag]

It took me a lot of time to compile this mex, so I'm sharing my experience here to spare people what I went through.

My configuration:
MATLAB 2011a running on a 32-bit Windows 7

How to compile the FORTRAN file:
================================
1- Download GNUMEX
sourceforge.net/projects/gnumex/

2- Download and install MINGW (as required by GNUMEX)

3- Copy mexinterface_c.f90 from gnumex/examples to zipSUSAN

4- Modify nlDiffFilter.f90 as follows:
a. Replace
use mexf90
with
use mexinterface

b. Replace
A => mxGetPr(prhs(1))
with
call c_f_pointer(mxGetPr(prhs(1)),A)

c. Replace
table => mxGetPr(prhs(2))
with
call c_f_pointer(mxGetPr(prhs(2)),table)

d. Replace
mask => mxGetPr(prhs(3))
with
call c_f_pointer(mxGetPr(prhs(3)),mask)

e. Replace
B => mxGetPr(plhs(1))
with
call c_f_pointer(mxGetPr(plhs(1)),B)

5- In MATLAB navigate to GNUMEX folder (as instructed in GNUMEX's readme) and run gnumex from MATLAB's prompt. (configure the compiler to be gfortran).

6- Navigate to zipSUSAN and issue the following 2 commands:
mex -c mexinterface_c.f90
mex nlDiffFilter.f90

Done.
================================

Now run demoSUSAN :)