Institute of Information Theory and Automation

Image fusion methods for degraded and incomplete data

Project leader: Prof. Ing. Jan Flusser, DrSc.
Department: ZOI
Supported by (ID): GA102/00/1711
Duration: 2000 - 2002
Details: here
Publications at UTIA: list

Abstract:

The project is concerned with automatic fusion of digital images. The input images are assumed to be blurred by unknown linear filters, geometrically distorted and corrupted by additive noise. By image fusion we understand the process of extracting useful information from each individual input frame and its integration into one output image, the quality of which is of higher level than that of the inputs. In this project, we propose to carry out the fusion in two stages: (a) geometric registration of the input frames and (b) multiframe blind deconvolution. Since the current registration techniques were proven to fail when applied to blurred images, we want to develop a new method especially for this purpose. This method is assumed to be based on our recently derived invariants to blurring. In the second stage, we want to propose a new method for multiframe blind deconvolution, which should be more numerically stable and faster than the previously published ones.
Responsible for information: ZOI
Last modification: 07.10.2009
Institute of Information Theory and Automation