The procedure (for metals with dislocation cell structure) you are looking for is described in Acta Metallurgica, 32, 3, 1984, Pages 333–342(T. Ungar, H. Mughrabi, D. Rönnpagel, M. Wilkens). It describes both the experimental procedure in which instrumental broadening is minimized through an analyzer crystal, and the theoretical treatment in which the X-ray line profile is analyzed by separating contribution from "cell walls" and "cell interior". The relationship between the half width parameters of the (essentially fitted) peak profiles and dislocation densities is described in the paper.
Crystals with low dislocation densities are harder to measure. You will need to step up the experimental side by further reducing instrumental broadening (e.g. arrangements of monochromator - sample -analyzer crystal).
As you can tell from my above statements I am a skeptic with respect to procedures where instrumental broadening is not suppressed but rather "deconvoluted" from the line profile of your samples. Errors are large, and the method in my opinion is not suitable for small dislocation densities.
Thank you very much for your suggestions. I think what you have suggested is used for polycrystalline samples. However, I am interested in single crystals like BGO, CdWO3, LiNbO3 etc using HRXRD rocking curve.