Description of slgaborbands

Home > sltoolbox > imgproc > slgaborbands.m

slgaborbands

PURPOSE

SLGABORBANDS Generates a set of Gabor kernels

SYNOPSIS

function FB = slgaborbands(w, scales, orientations)

DESCRIPTION

SLGABORBANDS Generates a set of Gabor kernels

 $ Syntax $
   - FB = slgaborbands(w, scales, radians)

 $ Arguments $
   - w:                   the kernel window size is wxw
   - scales:              the scales (number of scales is m)
   - orientations:        the orientations (number of orientations is o)
   - FB:                  the array of filter banks

 $ Description $
   - Generates a set of complex filter banks for Gabor, with the 
     scales and orientations of the filter specified. If there are
     m scales and n orientations, it will product an w x w x m x n
     array containing m x n kernels.

 $ References $
   - C. Liu and H. Wechsler, ''Gabor Feature Based Classification Using the
   Enhanced Fisher Linear Discriminant Model for Face Recognition'', IEEE
   Trans. on Imag. Proc, Vol. 11, No. 4, Apr, 2002.

 $ History $
   - Created by Dahua Lin on Feb 25th, 2006.
   - Modified by Dahua Lin on Aug 4th, 2006.

CROSS-REFERENCE INFORMATION

This function calls:

This function is called by:

SUBFUNCTIONS

• function M = single_gabor_kernel(w, v, u)

SOURCE CODE

0001 function FB = slgaborbands(w, scales, orientations)
0002 %SLGABORBANDS Generates a set of Gabor kernels
0003 %
0004 % $ Syntax $
0005 %   - FB = slgaborbands(w, scales, radians)
0006 %
0007 % $ Arguments $
0008 %   - w:                   the kernel window size is wxw
0009 %   - scales:              the scales (number of scales is m)
0010 %   - orientations:        the orientations (number of orientations is o)
0011 %   - FB:                  the array of filter banks
0012 %
0013 % $ Description $
0014 %   - Generates a set of complex filter banks for Gabor, with the
0015 %     scales and orientations of the filter specified. If there are
0016 %     m scales and n orientations, it will product an w x w x m x n
0017 %     array containing m x n kernels.
0018 %
0019 % $ References $
0020 %   - C. Liu and H. Wechsler, ''Gabor Feature Based Classification Using the
0021 %   Enhanced Fisher Linear Discriminant Model for Face Recognition'', IEEE
0022 %   Trans. on Imag. Proc, Vol. 11, No. 4, Apr, 2002.
0023 %
0024 % $ History $
0025 %   - Created by Dahua Lin on Feb 25th, 2006.
0026 %   - Modified by Dahua Lin on Aug 4th, 2006.
0027 %
0028 
0029 
0030 m = length(scales);
0031 n = length(orientations);
0032 FB = zeros(w, w, m, n);
0033 
0034 for i = 1 : m
0035     for j = 1 : n
0036         FB(:,:,i,j) = single_gabor_kernel(w, scales(i), orientations(j));
0037     end
0038 end
0039 
0040 function M = single_gabor_kernel(w, v, u)
0041 % v - scale, u - orientation
0042 
0043 mincoord = fix(- w / 2);
0044 coords = mincoord + (0 : w-1);
0045 [X, Y] = meshgrid(coords, coords);
0046 
0047 kmax = pi / 2;
0048 f = sqrt(2);
0049 sigma = 2 * pi;
0050 
0051 k_v = kmax / f^v;
0052 phi_u = pi * u / 8; 
0053 k_uv = k_v * exp(i * phi_u);
0054 
0055 F1 = (k_v^2) / (sigma^2) * exp(-k_v^2 * abs(X.^2 + Y.^2) / (2*sigma^2)) ;
0056 F2 = exp(i * (real(k_uv) * X + imag(k_uv) * Y)) - exp(-sigma^2/2);
0057 M = F1 .* F2;
0058 
0059 
0060 
0061 
0062 
0063

---

Generated on Wed 20-Sep-2006 12:43:11 by m2html © 2003