Optimization-Based Fusion - Hyperspectral Image Fusion - page 100

7.2 Image Quality
Various aspects of image quality are useful to objectively quantify how good or
well-perceived the image is. These aspects can be different depending on the context
and requirements of the application. Let us look at some of the common aspects of
an image that are attributed to its quality.
1. Spatial Resolution : Spatial resolution of an image refers to the size of the smallest
object that can be detected or resolved. For digital images, the size of the smallest
resolvable object cannot be smaller than the pixel size. Therefore, the spatial
resolution is limited by the pixel size. The resolution is mainly determined by
the instantaneous field of view (FOV) of the imaging sensor which quantifies
the area of the scene viewed by the single sensor element. Images are said to
have fine or high resolution when each of the sensor elements sees a small area
of the scene, and thus, smaller objects can be easily discriminated. In coarse or
low resolution images, only large features are clearly visible. However, the terms
high or low, and fine or coarse are relative to the subject and the context. They
do not have unified definitions. In the case of remote sensing images, the spatial
resolution refers to the smallest area on the earth's surface being seen , or captured
by the single sensor element. The altitude of the imaging device, i.e., its distance
from the target area is an important factor in determining the spatial resolution
which differentiates remote sensing images from other ones. Since the sensors
are typically located at a high altitude, they can capture a large geographical area,
but cannot provide minor spatial details.
Images with higher spatial resolution are always desirable since they have a better
visual appearance for human observers. Also, higher information content is useful
for processing of machine vision and pattern classification algorithms. The tech-
nique of enhancing the spatial resolution of the image beyond what is provided by
the imaging sensor with simultaneous reduction or elimination of aliasing or blur-
ring is known as super-resolution [30]. Most of the super-resolution techniques
combine the information from multiple low resolution observations of the scene
to generate a single high resolution image, while single-frame techniques process
only a single low-resolution observation for the purpose of super-resolution. Var-
ious approaches to single-frame and multi-frame super-resolution can be found
in [17, 30, 31, 123, 127].
In the case of remote sensing images, the need for increasing spatial resolution
is often felt in the case of multispectral images. This is achieved by integrating
information from a panchromatic image having a higher spatial resolution. This
process, known as pan-sharpening has been discussed in Sect. 2.1.2 . However,
the pan-sharpening process is not yet found to be much relevant for hyperspectral
images.
2. Gray-Level Resolution : The radiometric or gray-level resolution is defined as
the smallest discriminable change in the gray level [65]. It is dependent on the
signal to noise ratio (SNR) of the imaging device. However, in digital images,
this factor is limited by the number of discrete quantization levels used to digitize