Imaging molecular structures separated by distances of a few nanometers still represents a complex challenge. Moreover, it is normally restricted to observations on thin (few micrometers) samples. In this work, we rotate the polarization of the excitation beam of two-photon excited fluorescence (TPEF) images to show that fluorescent structures at the molecular scale can be discriminated in a living organism. The polarization rotation generates a modulation of the signal intensity in each pixel of the TPEF images that carry information related to the fluorophore orientation. We analyze the signal modulation in every pixel of the polarization-resolved (PR) TPEF images through a Fourier analysis and generate images for the different Fourier components. Doing that, we show that two fluorophores oriented in different directions can be distinguished. Although by imaging the Fourier components the resolution of the optical system restricts the exact localization of two close molecules, discrimination is still possible even when the molecules are located at sub-diffraction distances. We propose a model that predicts this behavior, and demonstrate it experimentally in the neurons of a living Caenorhabditis elegans nematode, where we distinguish the walls of an axon with a diameter below the objective resolution. Since the technique is based in TPEF, the method can be extended to deep tissue imaging and has potential applications in single molecule detection, biological sensors, or super-resolution imaging techniques.