A method for efficient detection of the dominant local rotational symmetry in grey level images of CBED patterns is described. The general approach is to define a local measure of rotational symmetry that transforms the symmetry detection problem to an optimization problem, and obtain the symmetric regions by an efficient global optimization algorithm. In this paper, we present the correlation function as the measure of rotational symmetry. The value depends on the center of the supporting region, its size, and the rotational angle. Genetic algorithm is taken to search the global maximum of the rotational symmetry. According to the unique characteristic of CBED patterns, some segmentation method is applied to the images of CBED patterns to reduce the search space of global optimization algorithm, thus this method is shown in experiments both effective and quick.