A rough non-uniform ZnSe thin film on a GaAs substrate is optically characterised using imaging spectroscopic reflectometry (ISR) in the visible, UV and near IR region, applied as a standalone technique. A global-local data processing algorithm is used to fit spectra from all pixels together and simultaneously determine maps of the local film thickness, roughness and overlayer thickness as well as spectral dependencies of film optical constants determined for the sample as a whole. The roughness of the film upper boundary is modelled using scalar diffraction theory (SDT), for which an improved calculation method is developed to process the large quantities of experimental data produced by ISR efficiently. This method avoids expensive operations by expressing the series obtained from SDT using a double recurrence relation and it is shown that it essentially eliminates the necessity for any speed–precision trade-offs in the SDT calculations. Comparison of characterisation results with the literature and other techniques shows the ability of multi-pixel processing to improve the stability and reliability of least-squares data fitting and demonstrates that standalone ISR, coupled with suitable data processing methods, is viable as a characterisation technique, even for thin films that are relatively far from ideal and require complex modelling.