Page 239 - Digital Analysis of Remotely Sensed Imagery

CHAPTER 6

Image Enhancement

mage enhancement refers to data processing that aims to increase
the overall visual quality of an image or to enhance the visibility
Iand interpretability of certain features of interest in it. During
acquisition of remotely sensed imagery, the potential range of pixel val-
ues may not be fully utilized in recording the data owing to the atmo-
spheric effect and the limitations of the sensing system. Consequently,
the obtained data may have a poor quality, such as a low contrast, an
overly dark tone, or much radiometric noise. Eradication of such prob-
lems lies in image enhancement that may be carried out either nonspa-
tially, based on histogram information, or spatially within an operating
window. In nonspatial image enhancement, the output value of a pixel
is based solely on its input value without taking its neighboring pixels
into consideration. Namely, the value a pixel receives in the output
image is not affected by the value of its neighboring pixels. In spatial
image enhancement, the output value of a pixel is affected by that of
surrounding pixels within the operating window. Both spatial and
nonspatial enhancements are undertaken either for a single band or for
multiple bands. No matter how many spectral bands are involved, it is
worthwhile to note that image enhancement does not create any new
information in the output image. On the contrary, such processing is
usually accompanied by a loss of information. Thus, the enhanced
image may contain less information than the original image. As a mat-
ter of fact, it is the quality of the features of interest in the input image
that is enhanced at the expense of losing information about features of
no interest to the analyst. Whether the same pixel value is regarded as
information or noise depends utterly on the purpose of enhancing the
image.
This chapter on image enhancement consists of seven sections.
Covered in the first section are nonspatial image enhancement tech-
niques that include density slicing and contrast enhancement. This is
followed by spatial enhancement, such as spatial filtering, and edge
enhancement and detection. Afterward, the discussion shifts to mul-
tiple image manipulation in Sec. 6.5. The sixth part of this chapter is
devoted to image transformation. An example of principal compo-
nent analysis is provided to illustrate the undertaking of the
203