ABSTRACT As established in previous papers of this series, observables in highly distorted and magnified multiple images caused by the strong gravitational lensing effect can be used to constrain the distorting properties of the gravitational lens at the image positions. If the background source is extended and contains substructure, like star forming regions, which is resolved in multiple images, all substructure that can be matched across a minimum of three multiple images can be used to infer the local distorting properties of the lens. In this work, we replace the manual feature selection by an automated feature extraction based on SExtractor for Python and show its superior performance. Despite its aimed development to improve our lens reconstruction, it can be employed in any other approach, as well. Valuable insights on the definition of an ‘image position’ in the presence of noise are gained from our calibration tests. Applying it to observations of a five-image configuration in galaxy cluster CL0024 and the triple-image configuration containing Hamilton’s object, we determine local lens properties for multiple wavebands separately. Within current confidence bounds, all of them are consistent with each other, corroborating the wavelength-independence of strong lensing and offering a tool to detect deviations caused by micro-lensing and dust in further examples.