SUSANCORNER - Edge/corner detection or image smoothing.

Description
Syntax
Inputs
Outputs
References
See also
Function implementation

Description

Edge/corner detection or image smoothing using the SUSAN technique of [SB97] already implemented in [SUSAN].

Syntax

   Smap = SUSANCORNER(I);
   [Smap, Spt] = SUSANCORNER(I, 'c');
   Smap = SUSANCORNER(I, mode, 'Property', propertyvalue, ...);

Inputs

I : an input image with size (X,Y,C), where C>1 when I is multichannel.

mode : optional string for chosing the computing mode, on which the outputs of the SUSAN filter will depend; it is either:

'e' for detecting edges,
'ei' for edges overlaid on input image,
's' for smoothing mode,
'c' for extracting corners - output a (n,2) corners matrix with the coordinates of the n corners in the image,
'ci' for corners overlaid on the input image;

default: mode='e', ie the edges are extracted.

Property [propertyname propertyvalues]: *'thres' : optional parameter for brightness threshold, used by all modes; default: |thres=20;

'md' : distance mask, either a threshold (with 's' mode) or the string 'flat' for flat (3,3) masks (with 'e' and 's' modes).

'n' : optional boolean for avoiding the post-processing on the binary edge map (runs much faster) with 'e' modes; default: n=false;

'q' : optional boolean for faster (and usually stabler) 'c' modes; edge-like corner suppression not carried out; default: q=true.

Outputs

Smap : depending on the computing mode selected (see above), it can be either:

a logical (binary) array (mask); this is the default type used with the 'e' mode,
an image of same size as I when either the 's' mode or the 'ei' and 'ci' modes are selected (type uint8 with overlaid features in the latter cases),
a (n,2) corners matrix with the coordinates of the n corners in the image.

Spt : coordinates of the corner when mode='c'.

References

[SB97] S.M. Smith and J.M. Brady: "{SUSAN} - A new approach to low level image processing", International Journal of Computer Vision, 23(1):45-78, 1997. http://www.lems.brown.edu/vision/courses/image-processing/Readings/smith95susan.pdf

[SUSAN] Source code available at http://www.fmrib.ox.ac.uk/~steve/susan/

Function implementation

function [Smap, varargout] = susancorner(I,varargin)

parsing and checking parameters

error(nargchk(1, 18, nargin, 'struct'));
error(nargoutchk(1, 2, nargout, 'struct'));

if ~isnumeric(I)
    error('susancorner:inputerror','matrix required in input');
end

p = createParser('SUSANCORNER');
% mandatory parameter
% optional parameters
p.addOptional('mode', 'e', @(x)ischar(x) && ...
    any(strcmpi(x,{'e','ei','c','ci','s'})));
p.addParamValue('thres', 20, @(x)isscalar(x) && x>=0);
p.addParamValue('md', 'flat', @(x)(ischar(x) && strcmp(x,flat)) || ...
    (isscalar(x) && x>0));
p.addParamValue('n', false, @(x)isscalar(x) && (x==true||x==false));
p.addParamValue('q', true, @(x)isscalar(x) && (x==true||x==false));

% parse and validate all input arguments
p.parse(varargin{:});
p = getvarParser(p);

main computation

C = size(I,3);

if strcmpi(p.mode,'c') && nargout==2
    [Smap, varargout{1}] = susancorner_base(I, p.mode, p.thres, p.md, p.n, p.q);
else
     Smap = susancorner_base(I, p.mode, p.thres, p.md, p.n, p.q);
end

display

if p.disp
    figure,
     if any(strcmpi(p.mode,{'e','¨ei','s'})),
         imagesc(Smap);
     else
         tmp = false(size(I));
         for c=1:C,
             tmp(Smap{c}(:,1) + X*(Smap{c}(:,2)-1) + X*Y*(c-1)) = true;
         end
         imagesc(tmp);
     end
        axis image off;
        if C==1, colormap gray;  end;
        title('corners detected with SUSAN');
end

end % end of susancorner