Numerous butter features, such as spreadability, depend on the crystallization, which is observable on microscopical images of milk fat. As microstructures often reveal rough or irregular aspect, fractal tools appear to be particularly adapted to their study. In this paper, fractal techniques quantifying the texture of the image are applied to separate type, plant and day of ripening of different butters. The images are taken from butters produced in industrial conditions. The influence of the crystallization on the fractal parameters has been observed. Fractal quantifiers have demonstrated promising correlations with the day of ripening of the butter. Moreover, they have also shown to be efficient in separating different kinds of butters.