A FAST AND ROBUST FRAMEWORK FOR IMAGE FUSION AND ...

Chapter 1
Introduction
On the path to designing high resolution imaging systems, one quickly runs into the problem
of diminishing returns. Specifically, the imaging chips and optical components necessary to
capture very high resolution images become prohibitively expensive, costing in the millions of
dollars for scientific applications [5]. Hence, there is a growing interest in multi-frame image
reconstruction algorithms that compensate for the shortcomings of the imaging systems. Such
methods can achieve high-quality images using less expensive imaging chips and optical com-
ponents by capturing multiple images and fusing them. The application of such algorithms will
certainly continue to proliferate in any situation where high quality optical imaging systems
cannot be incorporated or are too expensive to utilize.
A block diagram representation of the problem in hand is illustrated in Figure 1.1,
where a set of images are captured by a typical imaging system (e.g. a digital camcorder).
As the relative motion between the scene and the camera, the readout noise of the electronic
imaging sensor (e.g. the CCD), and possibly the optical lens characteristics change through
the time, each estimated image captures some unique characteristic of the underlying original
image.
In this thesis, we investigate a multi-frame image reconstruction framework for fusing
the information of these low-quality images to achieve an image (or a set of images) with higher
1