D = VL_SIFTDESCRIPTOR(GRAD, F) calculates the SIFT descriptors of the keypoints F on the pre-processed image GRAD. GRAD is a 2xMxN array. The first layer GRAD(1,:,:) contains the modulus of gradient of the original image modulus. The second layer GRAD(2,:,:) contains the gradient angle (measured in radians, clockwise, starting from the X axis -- this assumes that the Y axis points down). The matrix F contains one column per keypoint with the X, Y, SIGMA and ANGLE parameters.

In order to match the standard SIFT descriptor, the gradient GRAD should be calculated after mapping the image to the keypoint scale. This is obtained by smoothing the image by a a Gaussian kernel of variance equal to the scale of the keypoint. Additionaly, SIFT assumes that the input image is pre-smoothed at scale 0.5 (this roughly compensates for the effect of the CCD integrators), so the amount of smoothing that needs to be applied is slightly less. The following code computes a standard SIFT descriptor by using VL_SIFTDESCRIPTOR():

  I_       = vl_imsmooth(im2double(I), sqrt(f(3)^2 - 0.5^2)) ;
mod      = sqrt(Ix.^2 + Iy.^2) ;
ang      = atan2(Iy,Ix) ;
grd      = shiftdim(cat(3,mod,ang),2) ;
grd      = single(grd) ;
d        = vl_siftdescriptor(grd, f) ;

Remark

The above fragment generates results which are very close but not identical to the output of VL_SIFT() as the latter samples the scale space at finite steps.

Remark

For object categorization is sometimes useful to compute SIFT descriptors without smoothing the image.

Options:

Magnif [3]

Magnification factor (see VL_SIFT()).

NormThresh [-inf]

Set the minimum l2-norm of the descriptors before normalization. Descriptors below the threshold are set to zero.