Page 49 - Digital Analysis of Remotely Sensed Imagery

22 Cha pte r O n e

resolution. Since the amount of energy emitted by targets is much
smaller than what is reflected, it is more difficult to achieve the same
spatial or radiometric resolution for images acquired over the thermal
infrared portion of the spectrum than over visible and near infrared
wavelengths.

1.6 Organization of the Book
This book is divided into 14 chapters. Chapter 2 comprehensively
surveys the main characteristics of existent remote sensed data
available for digital analysis. Also included in this chapter is how to
convert existing analog remote sensing materials into digital format
via scanning. Chapter 3 presents various media for storing remote
sensing data, and the common image formats for saving remote sensing
imagery and processed results. Also covered in this chapter are methods
of data compression, both lossy and error free. Contained in Chap. 4 is
a critical overview and assessment of main digital image analysis
systems, their major features and functions. A few of the lead players
are described in great depth, with the strengths and limitations of each
system critically assessed. How to prepare remote sensing data for
digital analysis geometrically forms the content of Chap. 5. After the
fundamentals of image geometric rectification are introduced, several
issues related to image rectification are addressed through practical
examples. Also featured in this chapter are the most recent developments
in image georeferencing, such as image orthorectification and real-time
georeferencing. Chapter 6 is devoted to image enhancement methods,
ranging from simple contrast manipulation to sophisticated image
transformation. Most of the discussion centers around processing in
the spectral domain while image enhancement in the spatial domain
is covered briefly.
Covered in Chaps. 7 to 11 are five unique approaches toward
image classification. Chapter 7 on spectral image classification begins
with a discussion on the requirements and procedure of image
classification. The conventional per-pixel-based parametric and
nonparametric methods, namely, unsupervised and supervised
methods, are presented next. Three supervised image classification
algorithms are introduced and compared with one another in terms
of their requirements and performance. This is followed by more
advanced classification methods, including subpixel and fuzzy
image classification. This chapter ends with a brief discussion on
postclassification processing.
With the advances in machine learning, new methods have been
attempted to perform image classification in the hope of achieving
higher accuracies. Two attempts of neural network classification and
decision tree classification form the focus of Chaps. 8 and 9, respectively.
After the various types of neural network structures are introduced in
Chap. 8, the discussion then shifts to network configuration and