SLIDEHISTOFUN - Local histogram-based image statistics.

Contents

Description

Function for computing local statistical features based on local histograms computed over square sliding windows.

Syntax

  O = SLIDEHISTOFUN(I);
  O = SLIDEHISTOFUN(I, func, win, bin, method);

Inputs

I : input image with size (X,Y,C) (C>1 for multispectral images) over which local statistics are computed.

func : function handle or cell of function handles to apply on the local histograms.

win : (optional) size of the local (square) sliding window over which statistics are computed; default win=3.

bin : (optional) number of bins used in the histogram representation as resolution reduction; default: bin is the max number of values available in the input image.

method : (optional) string indicating the method choose for computing the local histograms in sliding windows; it is either:

note that in the two first mentioned cases, the integral histograms need to be computed and stored prior to the function application; default: method='dist';

Output

O : an image of size (X,Y) when a single function is computed over the local histograms (func passed as a function handle, see above), or a cell of such images when several functions are estimated (func passed as a cell of functions handles); O(x,y) (or O{i}(x,y)) is the output of the function func (or func{i}) of the histogram computed on the local window centered at (x,y).

References

[Pori05] F. Porikli: "Integral histogram: a fast way to extract histogram features", Proc. IEEE CVPR, pp. 829-836, 2005. http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=1467353

[ZLLG10] K. Zhang, G. Lafruit, R. Lauwereins and L. Van Gool: "Joint integral histograms and its application in stereo matching", Proc. IEEE ICIP, pp. 817-820, 2010. http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=5653410

[SDH08] M. Sizintsev, K.G. Derpanis, A. Hogue: "Histogram-based search: a comparative study", Proc. IEEE CVPR, pp. 1-8, 2008. http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=4587654

See also

Ressembles: BLKPROC, COLFILT. Requires: SLIDEHISTOFUN_BASE.

Function implementation

function O = slidehistofun(I, varargin)

parsing parameters

error(nargchk(2, 13, nargin, 'struct'));
error(nargoutchk(1, 1, nargout, 'struct'));

if ~isnumeric(I)
    error('slidehistofun:inputparameter','vector or matrix required in input');
end

p = createParser('SLIDEHISTOFUN');
p.addRequired('func', @(x)iscell(x) || ...
    strcmpi(class(x),'function_handle'));  % has to be entered!!!
p.addOptional('win', 3, @(x)isscalar(x) && x>=1);
p.addOptional('bin', [], @(x)isempty(x) || ...
    (isscalar(x) && x>=1 && round(x)==x));
p.addOptional('method', 'dist', @(x)ischar(x) && ...
    any(strcmpi(x,{'dist','int','joint'})));

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

checking variables

if isempty(p.bin),  p.bin = max(I(:)) + 1;  end

if mod(p.win,2)==0,  p.win = p.win + 1;  end

main computation

O = slidehistofun_base(I, p.method, p.func, p.win, p.bin);

display

if p.disp
    if iscell(O)
        N = numel(O);
        figure; ndisp = 3; mdisp = ceil(N/ndisp);
        for i=1:N
            subplot(mdisp, ndisp, i),  imagesc(rescale(O{i})), axis image off;
            title(func2str(p.func{1}));
        end
    else
        figure, imagesc(rescale(O)), axis image off, title(func2str(p.func));
    end

end

end % end of slidehistofun