|M.Sc Student||Zaretsky Alexey|
|Subject||Resolution Enhancement of Endoscopic Images|
|Department||Department of Electrical Engineering||Supervisor||Professor Emeritus Arie Feuer|
Endoscopic imaging is a very important tool in modern medicine. Physicians set very high demands for image quality, but that quality is limited by endoscope fiber technology.
This research addresses the problem of resolution enhancement of endoscopic images. A mathematical model of an endoscope was developed, describing an endoscope system as a Linear Space Invariant (LSI) filtering with sampling. The problem of image resolution improvement was considered as a Super Resolution (SR) problem. Multi-channel decomposition method, based on well-known Papoulis theorem, was used for image resolution reconstruction. A bank of parallel independent filters, prior to sampling, was applied to an input image, creating additional information for SR. An input image is decomposed to channels by the filter-bank. In order to restore the original image resolution, the data of each sampled channel is filtered by respective reconstruction filter, and all channels are then summed. An efficient algorithm for calculating reconstruction discrete filters is presented, handling a general sampling grid. The special case of hexagonal sampling due to the geometry of the fiber bundle of the endoscope was considered.
Finding the proper family of filters for the parallel filter-bank is another important subject that we have dealt with. The filters must satisfy certain conditions, stated in the Papoulis theorem. In order to generate the data needed for the resolution enhancement we consider several filter families, which can be implemented by the optical system, located at the distal tip of the endoscope.