TENSANIFILT - Anisotropic Gaussian filtering of multispectral images.

Contents

Description

Perform Gaussian adaptive filtering of an image using an anisotropic tensor defined directly from the image, e.g. the Gradient Structure Tensor (GST) or the Hessian.

Syntax

   F = TENSANIFILT(I);
   [F, S, T] = TENSANIFILT(I, method, rho, sigma, ...
                            'Property', propertyvalue, ...);

References

[Weick97] J. Weickert: "Coherence-enhancing diffusion of colour images", Proc. of PRIA, vol. 1, pp. 239-244, 1997. http://www.sciencedirect.com/science/article/pii/S0262885698001024

[KMS00] R. Kimmel, R. Malladi and N. Sochen: "Images as embedded maps and minimal surfaces: movies, color, texture, and volumetric medical images", International Journal of Computer Vision, 39(2):111-129, 2000. http://www.springerlink.com/content/u3m2629171491109/

[MN02] M. Middendorf and H.-H. Nagel: "Empirically convergent adaptive estimation of grayvalue structure tensors", Proc. of DAGM, LNCS 2449, pp. 66-74, 2002. http://www.springerlink.com/content/fbbrynkyy8yhfd4q/

[SLS01] A.F. Sole, A. Lopez and G. Sapiro: "Crease enhancement diffusion", Computer Vision and Image Understanding, 84(2):241-248, 2001. http://www.sciencedirect.com/science/article/pii/S1077314201909452

See also

Ressembles: CONVOLUTION, TENSCALEDIRFILT, ADAPTIVEFILT, GEODESICFILT, MDLFILT. Requires: TENSANIFILT_BASE.

Function implementation

function [F, S, T] = tensanifilt(I,varargin)

parsing parameters

error(nargchk(1, 32, nargin, 'struct'));
error(nargoutchk(1, 3, nargout, 'struct'));

% mandatory parameter
if ~isnumeric(I)
    error('tensanifilt:inputerror','a matrix is required in input');
end

% optional parameters
p = createParser('TENSANIFILT');   % create an instance of the inputParser class.
p.addOptional('method', 'mid', @(x)ischar(x) && ...
    any(strcmpi(x,{'gst','gstort','wei','weickert','kim','kimmel','sol','sole',...
    'mid','middendorf','tsc','tschumperle'})));
p.addOptional('rho', 1, @(x)isscalar(x) && isfloat(x) && x>=0);
p.addOptional('sigma', 1, @(x)isscalar(x) && isfloat(x) && x>=0);
% p.addParamValue('scales',0.5:0.1:4, @(x)nb_dims(x)==1 && length(x)>=2);
% additional optional parameters
p.addParamValue('der', 'fast', @(x)islogical(x) || (ischar(x) && ...
    any(strcmpi(x,{'matlab','vista','fast','conv','fleck', ...
    'tap5','tap7','sob','opt','ana'}))));
p.addParamValue('int', 'fast', @(x)islogical(x) || (ischar(x) && ...
    any(strcmpi(x,{'matlab','conv','fast','ani'}))));
p.addParamValue('samp', 1, @(x)isscalar(x) && round(x)==x && x>=1 && x<=5);
% parameters for method [Weick97] and [KMS00]
p.addParamValue('a', 0.5, @(x)isscalar(x) && x>0 && x<1);
p.addParamValue('c', 1, @(x)isscalar(x) && x>0);
% parameters for method [SLS]
p.addParamValue('alpha', 1, @(x)isscalar(x) && x>=0);
p.addParamValue('beta', 1, @(x)isscalar(x) && x>=0);
p.addParamValue('eps', 0.5, @(x)isscalar(x) && x>0 && x<1);
% parameter for method [Tschumperle]
p.addParamValue('p1', 1, @(x)isscalar(x) && x>=0);
p.addParamValue('p2', 2, @(x)isscalar(x) && x>=0);

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

main computation

[F, S, T] = tensanifilt_base(I, p.method, p.rho, p.sigma, p.der, p.int, ...
    p.samp, p.a, p.c, p.alpha, p.beta, p.eps, p.p1, p.p2);

display

if p.disp
    % Display the tensor fields. The color is proportional to the size of the
    % tensor.
    U = perform_tensor_mapping(T,+1);
    U(:,:,1) = perform_histogram_equalization(U(:,:,1), 'linear');
    T1 = perform_tensor_mapping(U,-1);
    options.sub = 5;   plot_tensor_field(T1, I, options);
    figure, imagesc(rescale(F)), axis image off, title('anisotropic filtering');
    if size(F,3)==1,  colormap gray; end
end

end % end of tensanifilt