Filter Design for Image Decomposition and Applications to Forensics

Employing image filters in image processing applications, essentially matrix convolution operators, has been an active field of research since a long time, and it is so very much still today. In the first part, we give a brief overview of imaging methods with emphasis on applications in fingerprint recognition and shoeprint forensics. In the second part, we propose a generalized discrete scheme for image decomposition that encompasses many of the existing methods. Due to its generality, it has the potential to learn, for specific use cases, a highly flexible set of imaging filters that are related to one another by rather general conditions.