One dimensional frequency decomposition of oriented local images applied to fingerprints
| Title | One dimensional frequency decomposition of oriented local images applied to fingerprints |
|---|---|
| Summary | Study and develop 1D continuous frequency fitting of a predetermined number of (oriented) sinusoids to images which are strongly oriented locally, in particular forensic fingerprints. |
| Keywords | |
| TimeFrame | |
| References | |
| Prerequisites | |
| Author | Josef Bigun, Anna Mikaelyan |
| Supervisor | Josef Bigun |
| Level | Master |
| Status | Open |
One dimensional frequency decomposition arises in many applications. An important application, for which a strong theory and practice has been developed, is communication between a mobile-phone base station and the mobile-phones within its coverage. The base station must decode the individual signals of each mobile-phone even if several of them transmit messages concurrently. Source separation is done by frequency content estimation of the totality of the arriving signal. Two methods have been seminal in published studies, so called MUSIC and ESPRIT.
Images are two dimensional, so that the developed algorithms for 1D will not be directly suitable. Instead we will attempt to estimate the 1D frequency component of local images in the direction given by its dominant orientation, which in turn is estimated using the structure tensor theory. This should yield frequency parameters that are necessary for forensic fingerprint enhancement. Additional applications are possible for other types of images than fingerprints, e.g. local denoising.
Solid knowledge in Image Analysis and Mathematics is a prerequisite.
