DTV is image registration algorithm which was described in (Li, Y. 2015) -see the pic- in 3 steps:
Matlab code:
function TV = DTV (r_img,f_img)
%% find gradient domain for the images-> (forward finite difference)
[rx ry] = imgradientxy(r_img,'intermediate');
[fx fy] = imgradientxy(f_img,'intermediate');
%% L1-norm of x and y gradient
gradR_r = sum(sum((abs(rx) + abs(ry))));
gradR_f = sum(sum((abs(fx) + abs(fy))));
%% TV of residual image r in gradient domain
TV = sum(abs(gradR_r(:) - gradR_f(:))); %eq. 1
end
The DTV function doesn't measure the similarity correctly when change one of the image transformation.
Article Robust image registration in the gradient domain