الفهرس 
Chapter (2) IR ImagingOnly 14 pages are availabe for public view
 |  | from | 115 |  |  |
| | | from | 115 | | |
Abstract
This thesis is mainly concerned with the resolution enhancement of
Infrared (IR) images. These images are of relatively low contrast and they
have few details due to the lack of light in the IR imaging process. IR
images are acquired based on emission of heat from objects and their
environment. Such characteristics affect the ability of detecting targets.
Two trends are being presented in the thesis for the resolution
enhancement of IR images. The first one is presented with its
mathematical model, which depends on Single Image Super Resolution
(SIMSR). The SIMSR considers benefits of the sparse representations of
Low Resolution (LR) and High Resolution (HR) patches of the IR
images. In this trend, a database is first generated off-line for patches
from other images. For these images, both LR and HR patches are
available and a feed-forward neural network is used to learn the relation
between these patches. After the learning process, patches of the LR IR
image are fed to the created model to get the corresponding HR patches.
A Minimum Mean Square Error (MMSE) estimator is used in the
prediction process. The other trend depends on Compressed Sensing
(CS). The CS is a signal processing technique for efficiently acquiring
and reconstructing a signal. The basic idea of this trend is to perform
some sort of smart compression of IR images through a CS process in
order to save the bandwidth over the communication channel and
facilitate, or even enable its transmission. At the receiver side, a CS
reconstruction process is performed to reconstruct the original IR images.
The CS reconstruction is an inverse problem that is solved with an
optimization technique. After the CS reconstruction, a post-processing
stage that is based on SIMSR is performed to eliminate compression
problems resulting in high quality IR images. The common thread
between the two presented trends is that they deal with a large amount of
IR data. Hence, they can be classified as big data processing techniques.