A new characterization method of the lattice fringe images of turbostratic carbons is proposed. This method is based on the computation of their orientation field without explicit detection of fringes. It allows meaningful insights into the material nanostructure and nanotexture at several scales, either qualitatively or quantitatively. The calculation of pairwise spatial statistics of the orientation field at short distance provides measurements of the coherence lengths along any direction, in particular along and orthogonally to the layers. These statistics also allow representing orientation coherence patterns typical of the observed nanostructure. At larger distances, the mean disorientation of the fringes is computed and information about the homogeneity of the sample is obtained. An experimental validation is carried out on various artificial images and an application to the characterization of four bulk turbostratic carbons is provided.