Aufgabenstellung

Reflections of unwanted scenes in images are not only a problem in photography but also scientific applications of thermal cameras. At ITS, we are investigating highly sophisticated cooling configurations for turbine blades under realistic conditions in complex test rigs. The surface temperature of the test objects are captured with an infrared camera to evaluate the performance of different cooling configurations. In these thermal images reflections from the test rig can be seen on the surface of interest, downgrading the quality of the measurements. This why in this thesis, different methods for the removal of these reflections shall be studied.

Different reflection removal techniques exist in literature: for single images, multiple images, using multi-focus approach, moving the camera etc. Many of these approaches are based on deep neural networks. These different techniques shall be used to solve the problem and to identify the best method for our problem to ensure high quality thermal measurements.

Knowledge in Python and Image Processing is beneficial but not mandatory.