The application of exit-plane wave function (EPWF) images for an atomic-scale compositional mapping of interfaces in the GaAs-AlAs system is investigated. Image simulations based on the multislice algorithm were performed to obtain EPWF images over a wide range of compositions and specimen thicknesses. The EPWF-amplitude images were further analyzed by the method of factorial analysis of correspondence. A simultaneous analysis of the effects both thickness and composition on image contrast yielded two eigenvectors representing the individual contributions from the two effects. This separation was however not complete, leaving a residual contribution from thickness on the composition-sensitive eigenvector and vice versa. Extraction of the chemical content from the composition-sensitive eigenvector is still possible by a parameterization of the residual thickness contribution. A simpler procedure for quantifying the composition across an interface is proposed based on the analysis of images of AlxGa(1-x)As with different Al/Ga composition but same specimen thickness.