CHAPTER ONE
1. Introduction to Digital Image Processing
1.1 Why do we collect remote sensing imagery?
Basically we actively collect earth resource data to test hypothesis and simulate or
model the environment. Thus, the data is collected using either in situ or remote
sensing methods. One method of in situ data collection is to have a researcher
located at the study area who is observing or questioning the phenomena under
investigation. It may also be performed using transducers that are placed in direct
contact with the phenomena of interest. Remote sensing, however, departs from
the aforementioned data collection methods because the sensor is remote from the
phenomena; that is, it is not in direct physical contact with it.
1.2 What type of data can be remotely sensed?
To answer this question it is necessary to identify two variables. First, there are
variables that remote sensors can measure directly. This means that the remotely
sensed data can provide fundamental biological and/ or physical (biophysical)
information directly without ancillary data. A good example is temperature mapping
using (thermal infrared), location (X, Y and Z) using stereoscopic photograph.
The second general group of variables that may be remotely sensed are hybrid
variables created systematically analyzing more than one biophysical variable. For
example, plant chlorophyll absorption characteristics, its temperature, and moisture
content to detect vegetation stress. Land cover mapping is also hybrid variable as it
requires evaluating several variables such as object color, location, temperature etc.
1.3 Resolution considerations
Resolution or resolving power is a measure of the ability of an optical system to
distinguish between signals that are spatially near or spectrally similar. The ability
to measure a biophysical variable using remote sensing requires careful consideration
of four types of resolution: spectral, spatial, temporal and radiometric resolutions.
1
Spectral resolution refers to the dimension and number of specific wave length
intervals in the electromagnetic spectrum to which a sensor is sensitive. The size of
the intervals or band may be large i.e., coarse, as with panchromatic black and white
aerial photography (0.4-0.7µm) or relatively small i.e., fine, as with band 3 of Landsat
5 thematic mapper sensor system (0.63-0.69µm).
Spatial resolution is a measure of the smallest angular or linear separation between
two objects that can be resolved by the sensor. For aerial photography it is the
number of resolved lines per millimeter on the image. For other sensor systems it is
simply the dimension of the ground projected instantaneous field of view (IFOV) of
the sensor system that moves across the terrain collecting information. A useful
rule here is that in order to detect a feature, the spatial resolution of a sensor
system should be less than half size of the feature measured in its smallest
dimension.
Temporal resolution refers to how often a given sensor obtains imagery of a
particular area. Ideally, the sensor obtains data repetitively to capture unique
discriminating characteristics of the phenomena of interest.
Radiometric resolution defines the sensitivity of a detector to differences in signal
strength as it records the radiant flux reflected or emitted from the terrain. It
defines the number of just discriminable signal levels, consequently, it may be a
significant element in the identification of scene objects. For example, MSS Landsat
1-5 6bits (0-63) then expanded data in 3bands in 7bit (0-127) with spatial resolution
of 79m; Landsat TM 4 and 5 in 8bits (0-255) at 30x30m in 6 of 7 bands.
1.4 Essence of processing remote sensing data digitally
Human beings are adept at visually interpreting images produced by certain remote
sensing devices, especially cameras. One could ask: why try to mimic or improve on
this capability? First, there are certain thresholds beyond which the human
interpreter cannot detect just noticeable differences in the imagery. For example,
it is commonly known that an analyst can discriminate only about 16 shades of gray
when interpreting continuous tone black and white aerial photography. If the data
were originally recorded with 256 shades of gray, there may be more subtle
information present in the image than the interpreter can extract visually.
2
Furthermore, it makes interpretations generally unrepeatable. Conversely, the
results obtained by computer are almost always repeatable (even when wrong). It
should be kept in mind that it is not possible to say that digital image processing is
superior to visual image analysis. Rather, there may be times when digital approach
is better suited to the problem at hand. But there exists several image analysis tasks
and basic element of image interpretation that manual and computer assisted
approaches share. First, both manual and digital analysis of remotely sensed data
seek to detect and identify important phenomena in the scene. Thus they have the
same goals however they may follow significantly different paths. Most of the
fundamental elements of image interpretation (size, shape, shadow, site, texture,
height, association and pattern) are used in manual image analysis but few are used
in computer assisted image processing (tone/color using fundamental statistical
pattern, and texture, some level of association).
1.5 Perspectives of analysis
Most often we are concerned with the art and science of applying digital image
processing techniques to remotely sensed data for the extraction of useful earth
resource information. The mathematical bases of the techniques are important, but
what is more important is how these techniques may be applied to real world
problems. Consequently, the digital image processing concepts and their
mathematical underpinnings will be presented in this course.
1.6 Overview of image processing considerations
Prior to extracting meaningful information from remote sensing data the image
analyst must evaluate several important image processing considerations. In short
the analyst has to go through the steps indicated in figure 1.1.
3
1. STATEMENT OF THE PROBLEM 6. PREPOROCESSING
Identify criteria Radiometric correction
Formulate hypothesis Geometric correction
2. DATA ACQUISITION 7. IMAGE ENHANCEMENT
Digitize or For further digital image
Collect or analysis
Purchase digital data For visual analysis
3. IMAGE PREPROCESSING SYSTEM 8. THEMATIC INFORMATION
Select or configure appropriate EXTRACTION
technology Identify criteria
Perform supervised or
unsupervised analysis
Evaluate accuracy
4. INITIAL STATISTICS EXTRACTION 9. GIS (Optional)
Compute univariate and multivariate Place thematic information in
statistics to asses image quality raster or polygon based GIS
Ask appropriate question
5. INITIAL DISPLAY 10. SOLVE PROBLEMS
Video or hard copy display to assess Accept or reject hypothesis
image quality
Figure 1. 1 Image considerations
