Photointerpretation and Remote Sensing Methodology

Deceti.jpg (13851 bytes)

Photointerpretation and Remote Sensing Methodology

3.1 Introduction

3.1.1 Analogue and Digital Sensors and Systems

3.2 Analogue and digital methods

3.2.1 Digital Image Analysis vs. Visual Interpretation
3.2.2 Advantages and disadvantages of analogue and digital methods

3.3 Pre-processing of digital Remotely Sensed Imagery

3.4 Simple Image Enhancement of Remotely Sensed Imagery

3.4.1 Contrast Enhancement
3.4.2 Pseudo-colour Enhancement - Density Slicing
3.4.3 Image enhancement through basic numerical calculations
3.4.4 Edge enhancement
3.4.5 Filtering
3.4.6 Special enhancement transformations

3.5 Classification

3.1 Introduction
Photointerpretation and Remote Sensing Methodology can be defined as: the dialectic and interdisciplinary integration of personal experience, reasoning, specific scientific knowledge and expertise and ground truth, in order to investigate, qualitatively and quantitatively, the objects, facts, characteristics, phenomena and patterns of the natural and socio economic reality, as well as of their multidimensional relations, interactions and interdependencies, and their change trends in time, by using their remotely sensed imagery.
3.1.1 Analogue and Digital Sensors and Systems	TOP

Artificial system of acquisition of analogue and digital remote sensing imagery based on the recording of the reflected electromagnetic radiation.

image2.jpg (12230 bytes)

Artificial system of remote sensing on the basis of the recording of the emitted (thermal) electromagnetic radiation.

image6.jpg (4192 bytes)

Active system of acquisition of remote sensing data on the basis of the transmittance of artificial EMR and the recording of the radiation backscattered from the earth.

image7.jpg (5370 bytes)

Source: Rokos “The contribution of remote sensing to the observation, monitoring and projection of the environment” in Environmental Crisis, ELKAM, Athens, 1993.

3.2 Analogue and digital methods	TOP
When processing and analysing remotely sensed images, manual and digital methods are usually combined.

image4.gif (60167 bytes)

In manual image processing and analysis, most of the fundamental photorecognition elements of visual interpretation, as well as their appropriate combinations, are used. Manual interpretation is often limited to analysing only a single channel of data or a single image or stereoscopic model at a time.
A human interpreter can only detect and evaluate noticeable differences in the imagery. Furthermore, he cannot carry out repeatable interpretation work.

Digital image processing and analysis requires a computer system, with the appropriate hardware and software to process the remotely sensed data, which is recorded in digital format.

image5.gif (9354 bytes)

Both analogue and digital approaches share several image analysis tasks and basic photorecognition elements of image interpretation:

Source : Estes et. al.

3.2.1 Digital Image Analysis vs. Visual Interpretation	TOP
a. Manual (visual) approach to image interpretation Remote sensing data can be basically seen as wavelength intensity information, which needs to be decoded before the message can be fully understood. This decoding process is analogous to the interpretation of the remotely sensed imagery, which relies on our knowledge of the properties of electromagnetic radiation. In order to extract meaningful information out of these data the image interpreter has to exercise his judgement, his scientific knowledge, his general knowledge of the phenomena as well as his experience, so that he will be able to make truthful assumptions about the object/feature under investigation. The first stage of image interpretation is known as detection. The detection stage is naturally followed by the recognition and identification stage in which the image interpreter has to exercise general, local, as well as specific levels of reference to allocate objects into known categories. The general level is the interpreter’s knowledge of the phenomena and processes to be interpreted, the local level is the interpreter’s intimacy with his own local environment, and the specific level is the interpreter’s deeper understanding of the processes and phenomena that he wants to interpret. In recognition and identification, the non-geometric image characteristics of tone or colour, texture, pattern, shape, shadow, size and location normally give clues. The result of identification is a list of objects and features in the area. These form the basis of delineation of areas having homogeneous patterns and characteristics. This is the analysis stage. Each delineated area has to be classified through a process of induction (general inference from particular cases) and deduction (particular inference from general observations). Accuracy is then controlled by field checks. The final stage of the interpretation is classification, producing spatial data which can be displayed as maps, or incorporated into a Geographic Information System.
b. Computer-assisted approach The manual approach suffers from its inability to deal quickly with a large quantity of image data. The development of computers led to the invention of “digital” methods of image interpretation. Image interpretation is basically a classificatory process identification and recognition can be treated in mathematical terms, provided that image data in digital form are available. However, the computer cannot replace the knowledge, experience, intelligence or understanding of the human – image interpreter. Various techniques such as Artificial Intelligence and Image Understanding, developed by scientists in order to improve the results of computer assisted image interpretation are being studied. The computer – assisted approach is usually based on:
statistical and syntactical pattern recognition,
the decision theoretic approach, and
symbolic reasoning.
Therefore, visual interpretation of digital imagery provided by remote sensing platforms does not allow full exploitation of the data provided. A human can only visually interpret 3 layers of remotely sensed information at a time. Further, our visual acuity does not allow us to identify all spectral differences in imagery. Machine processing of imagery allows for the quantitative analysis of all spectral bands in imagery simultaneously, and is able to detect subtle differences that we cannot.

3.2.2 Advantages and disadvantages of analogue and digital methods:
TOP

a. Analogue methods:
a1. Characteristics and advantages
Traditional: intuitive.
Simple, inexpensive equipment.
Uses brightness and spatial content of the image.
Usually single channel data or three channels at most.
Subjective, concrete, qualitative.

a2. Disadvantages:
The human interpreter understands much easier remote sensing data from the visible part of the spectrum and can observe and analyse only one image at a time. Therefore, the manual approach of image interpretation is recommended:

when the area under investigation (study) is rather small.

when the recognition and identification of an object/earth feature is necessary for the study of a specific problem/phenomenon.

when the “spectral signatures” of the object/earth feature under investigation are rather confusing and can not be easily distinguished.

when high spectral or spatial resolution is required in order to be able to distinguish the characteristics needed in our study.

b. Digital methods
Recent: require specialised training
Complex, expensive equipment.
Rely chiefly upon brightness and spectral content, limited spatial.
Frequent use of data from several channels.
Objective, abstract, quantitative.

b1. Advantages
Cost-effective for large geographic areas
Cost-effective for repetitive interpretations
Cost-effective for standard image formats
Consistent results
Simultaneous interpretations of several channels
Complex interpretation algorithms possible
Speed may be an advantage
Explore alternatives
Compatible with other digital data

b2. Disadvantages:
Expensive for small areas
Expensive for one-time interpretations
Start-up costs may be high
Requires elaborate, single-purpose equipment
Accuracy may be difficult to evaluate
Requires standard image formats
Data may be expensive, or not available
Preprocessing may be required
May require large support staff
Therefore, the computer-assisted approach is recommended:

when the area of concern is large.

when we are not interested in the detailed detection of objects/earth features.

when the area of concern is homogeneous and therefore the “spectral signatures” of object/earth features are easily distinguished.

when high spectral and spatial resolution are not particularly important for the results of our study.

Reference: R. Douglas Ramsey Doug@nr.usu.edu, Utah University, C.P.Lo., “Applied remote sensing”, University of Georgia.

3.3 Pre-processing of digital Remotely Sensed Images
TOP

During the digital remotely sensed images' acquisition, some “internal” or “external” errors degrade their quality and consequently the accuracy, the completeness and the reliability of the photointerpretation analysis that follows.
“Internal” errors are due to the remote sensors/systems, are systematic and may be determined by in-flight/before-flight calibration measurements.
“External” errors are due to many different platform perturbations that influence the modulation of remotely sensed image characteristics.
These non-systematic errors can be determined by relating points on the ground (known by their geodetic co-ordinates), which are mathematically correlated, with their well-defined positions on the remotely sensed images.
The basic errors during the remotely sensed images' acquisition, that degrade the images' quality and set certain restrictions to the possibilities of the Photointerpretation - Remote Sensing Methodology, are radiometric and geometric ones.
So, before the photointerpretation analysis and, more generally, before the remotely sensed images' exploitation, either only radiometric corrections, or only geometric ones or, and most commonly, both radiometric and geometric corrections are required.

a. Radiometric Corrections
The most common radiometric corrections of remotely sensed images acquired by multi-spectral scanners, are:

the restoration of the 6th line dropout ("loss of data"),

the restoration of the 6th line caused by the abnormality of the signal ("striping"),

the restoration of the horizontal lines parts displacement,

the restoration of the atmospheric diffusion and absorption results, which can eliminate the possibility of data acquisition from the explored region (for some areas of the electromagnetic spectrum). It has to be noted that the images acquired with wavelength λ > 0,7 μm (infrared), are practically clear of the atmospheric diffusion effects.

b. Geometric corrections
During the acquisition of digital multi-spectral remotely sensed images, some systematic and non-systematic errors influence the images' geometrical quality.
For systematic errors, relevant geometric corrections could be performed depending on:

the movement’s and the orbit’s elements of the carried platform or the remote system,

the element concerning the basic characteristics, the properties, the calibration measurements and the distortions of this particular remote system.

For non-systematic errors, such as:

errors resulting from the accidental altitude divergence of the remote system’s orbit (these errors cause a change of the remotely sensed images scale), and

errors resulting from accidental divergences of the remote system axis, from the specified places of reference (they are, usually, the first perpendicular, the second parallel to the flight axis and the third vertical to the two others previously mentioned),

is required the knowledge of the geodetic co-ordinates of some relating points on the ground, which, simultaneously, are well–defined on the remotely sensed images and consequently their image co-ordinates can be measured in pixels lines and columns.

Reference:
Rokos, D. “Special Applications of Photointerpretation and Remote Sensing”, NTUA, 1979.
Jensen, J. “Introductory digital image processing”, University of South Carolina, 1986.

3.4 Simple Image Enhancement of Remotely Sensed Images
TOP

In order to facilitate the processing of remotely sensed images, we often built image enhancement algorithms.
This can be achieved either by:

a. modification of each pixel regardless of the value of the adjacent pixels, or
b. modification of each pixel, taking into account the pixels around it.

The most common techniques in image enhancement are:

1. Contrast Enhancement.
2. Pseudo-colour Enhancement - Density Slicing.
3. Image enhancement through basic calculations.
4. Edge enhancement.
5. Filtering.
6. Other transformations such as: Principal Component Analysis, Vegetation Index, Kauth-Tomas/Tasseled Cap transformation, Multiple Discrimination Analysis, Hue-Saturation-Intensity Analysis, Fourier Analysis etc.

Additionally, image reduction and magnification can enhance image resolution and provide more particular information on the whole or a part of the image, as well as a histogram of pixel values along a line on the image.

3.4.1 Contrast Enhancement
TOP

Remote sensing systems record reflected and emitted energy from earth surface materials. Ideally, one material would reflect a tremendous amount of energy in certain wavelengths, while another material would reflect much less energy in the same wavelengths. This would result in contrast between the two types of materials when recorded by a remote sensing system. Unfortunately, different materials often reflect similar amounts of radiant flux throughout the visible and near-infrared portion of the electromagnetic spectrum, resulting in a relatively low contrast image.
In an image like this, all characteristics of earth (related to the natural and the socioeconomic reality and their changes through time etc.), become similar to the environment around them and not discernible.

3.4.2 Pseudo-colour Enhancement - Density Slicing
TOP

Pseudo-colour enhancements of an image intend to convert it from greyscale to colour (not necessarely to natural colours or like false infrared), in order to use all values of the electromagnetic spectrum.
The discrimination ability of the human eye is usually limited to 16 different grey levels. This limitation may be overcome if greyscale images are converted to pseudo-colour. This conversion allows the utilisation of the entire range of the image brightness values or portions of it in different hue values.
Density Slicing is the technique that assigns different hue values to each greylevel, in order to make the discrimination of different homogenous zones easier.
Usually, the selection of the successive portions is arbitrary and this causes loss of information. Studying the histogram and selecting the portion that refers to discernible homogenous zones could prevent this.
In this case the final image after density slicing control will be enhanced and much easier to interpret.

3.4.3 Image enhancement through basic numerical calculations
TOP

A remotely sensed image of the earth is a 3-D numerical matrix.

X, Y axes constitute a Cartesian reference system of spatial features.

Z-axis refers to the pixels' digital values for the amount of the energy emitted/reflected from every snapshot of earth surface.

The pixel’s digital value range is determined by the radiometric resolution of the sensor:
0 – 63 for radiometric resolution of the sensor 6 bits (IRS)
0 – 127 for radiometric resolution of the sensor 7 bits
0 – 255 for radiometric resolution of the sensor 8 bits (Landsat TM)
0 – 511 for radiometric resolution of the sensor 9 bits
0 – 1023 for radiometric resolution of the sensor 10 bits (NOAA)
0 – 2047 for radiometric resolution of the sensor 11 bits (Ikonos)
The amount of the radiation emitted/reflected from specific features on the earth surface depends not only on their physical, chemical or biological characteristics, but also on a series of other factors (terrain features, sun ratioing etc.).
This could cause difficulties during image interpretation, which may be eliminated by applying basic numerical calculation (subtraction, rationing) on the proper components (bands) of a multi-spectral digital remotely sensed image.
So, we may have image enhancement with the following characteristics:

a. Addition of brightness values
We can make an enhanced image by adding two or more remotely sensed images of the same date, time, region and dividing the sum with the number of the remotely sensed images.
This image is the average of its components and is released from the total noise allocation.

b. Subtraction of brightness values
By subtracting the appropriate two bands of two multi-spectral satellite images of the same area acquired at the same date in different years or in different seasons of the same year, we can view the annual/seasonal changes in the area, because of the nullification of the same pixel’s digital values.

c. Ratioing of Remotely Sensed Images
The image interpretation potential of a multi-spectral digital remotely sensed image is often bounded by the fact that specific features of one homogeneous zone reflect/emit electromagnetic radiation in different intensity. So, the features appear on the image in different grey tones. The terrain features cause this – humidity, seasonal change etc.
Ratioing the appropriate two components (bands) of the remotely sensed images, we can reduce the impact of the above mentioned factors.

3.4.4 Edge enhancement
TOP

Image interpretation of a remotely sensed image may be easier if the edges of the objects/characteristics are enhanced by an edge enhancement operation, so that their shapes and details are enhanced. Generally, what eyes see as pictorial edges are simply sharp defferences in brightness value between two pixels. The edge enhancement of a remotely sensed image can be performed, either with linear edge enhancement or with non-linear edge enhancement.

3.4.5 Filtering
TOP

The filtering of a remotely sensed image aims at the enhancement/improvement of the image, either with the elimination or compression of certain spatial frequencies and linear characteristics which abstract to the interpretation of other interesting characteristics (road network etc.) or with the enhancement of spatial frequencies and linear characteristics which concern us the most (boundaries of water resources).
The procedure, which is used to decompose a remotely sensed imagery to its different spatial frequency components, is Fourier Analysis.
Applying Fourier Transform on the original remotely sensed image is a way to enhance certain interesting spatial frequency components for image interpretation, contrary to others that do not interest us.
The algorithms, used for such an enhancement/improvement, are filters such as:

low pass filters

high pass filters.

High pass filters emphasize high frequency spatial characteristics, thus enhancing the linear characteristics of a remotely sensed image.
Low pass filters, on the other hand, emphasise low frequency spatial characteristics, causing a “smoothing” of the remotely sensed image.

3.4.6 Special enhancement transformations
TOP

Some of the special enhancement transformations used for natural resource inventories and monitoring, are related to principles, methods and techniques already mentioned.
So, we can use transformations, such as:

Principal Component Analysis of a remotely sensed image

Multiple Discriminant Analysis of a remotely sensed image.

Reference:
Rokos, D. “Special Applications of Photointerpretation and Remote Sensing”, NTUA, 1979.
Jensen, J. “Introductory digital image processing”, University of South Carolina, 1986.

3.5 Classification
TOP

Remote sensing data are records of reflected and emitted electromagnetic energy presented as picture like images. In order to extract meaningful information out of the data, an interpretation of the image has to be carried out. The image interpretation is made with the help of an electronic computer, where use of mathematical algorithms takes place, provided of course, that the data are in digital form.
The classification methods of the computer-assisted approach are:

Supervised classification: Representative sample sites of known land cover type (from a ground truth survey or from a map e.g. olives, forest, sea), called “training areas”, are selected in order to compile a numerical “interpretation key” that describes the spectral attributes of each feature type of interest. In other words these pixels from the “training areas” will be used to “train” the computer so that each pixel from the data set will be compared numerically to each category in the interpretation key and labeled with the category it “looks most like”. There are a number of numerical strategies (means, standard deviations etc.) that can be employed to make this comparison between unknown pixels and pixels derived from the “training areas”. The comparison is always made with the use of a specific classification algorithm, like the Gaussian Maximum Likelihood, the Parallelepiped and the Minimum distance to means algorithms.

Unsupervised classification: In this approach no training samples are used. The pixels from the remote sensing data are classified by the method of cluster analysis, which can identify natural groupings of patterns. The classes that result from unsupervised classification are spectral classes. Because they are based solely on the natural groupings in the image values, the identity of the spectral classes will not be initially known. The nature of each grouping is determined afterwards by ground truth surveys.

Unsupervised classification is not generally as effective as supervised classification because of the absence of training sets to control the results, especially when classes are only marginally separable. However, supervised classification is still too slow to handle a massive influx of satellite multi-spectral data. The results of the classification from the computer-assisted approach can be presented as a thematic map, which can be stored in digital form. The computer can also display numerical information on the area of the mapped classes or the frequency of the occurrence of each class and other useful statistical data if required.

Reference:
C.P.Lo., “Applied remote sensing”, University of Georgia.
Jensen, J. “Introductory digital image processing”, University of South Carolina, 1986.

TOP

National Technical University of Athens
Dept. of Rural & Surveying Engineering
Laboratory of Remote Sensing

3.2.2 Advantages and disadvantages of analogue and digital methods:	TOP
a. Analogue methods: a1. Characteristics and advantages Traditional: intuitive. Simple, inexpensive equipment. Uses brightness and spatial content of the image. Usually single channel data or three channels at most. Subjective, concrete, qualitative.
a2. Disadvantages: The human interpreter understands much easier remote sensing data from the visible part of the spectrum and can observe and analyse only one image at a time. Therefore, the manual approach of image interpretation is recommended:
when the area under investigation (study) is rather small.
when the recognition and identification of an object/earth feature is necessary for the study of a specific problem/phenomenon.
when the “spectral signatures” of the object/earth feature under investigation are rather confusing and can not be easily distinguished.
when high spectral or spatial resolution is required in order to be able to distinguish the characteristics needed in our study.
b. Digital methods Recent: require specialised training Complex, expensive equipment. Rely chiefly upon brightness and spectral content, limited spatial. Frequent use of data from several channels. Objective, abstract, quantitative. b1. Advantages Cost-effective for large geographic areas Cost-effective for repetitive interpretations Cost-effective for standard image formats Consistent results Simultaneous interpretations of several channels Complex interpretation algorithms possible Speed may be an advantage Explore alternatives Compatible with other digital data b2. Disadvantages: Expensive for small areas Expensive for one-time interpretations Start-up costs may be high Requires elaborate, single-purpose equipment Accuracy may be difficult to evaluate Requires standard image formats Data may be expensive, or not available Preprocessing may be required May require large support staff Therefore, the computer-assisted approach is recommended:
when the area of concern is large.
when we are not interested in the detailed detection of objects/earth features.
when the area of concern is homogeneous and therefore the “spectral signatures” of object/earth features are easily distinguished.
when high spectral and spatial resolution are not particularly important for the results of our study.
Reference: R. Douglas Ramsey Doug@nr.usu.edu, Utah University, C.P.Lo., “Applied remote sensing”, University of Georgia.

3.3 Pre-processing of digital Remotely Sensed Images	TOP
During the digital remotely sensed images' acquisition, some “internal” or “external” errors degrade their quality and consequently the accuracy, the completeness and the reliability of the photointerpretation analysis that follows. “Internal” errors are due to the remote sensors/systems, are systematic and may be determined by in-flight/before-flight calibration measurements. “External” errors are due to many different platform perturbations that influence the modulation of remotely sensed image characteristics. These non-systematic errors can be determined by relating points on the ground (known by their geodetic co-ordinates), which are mathematically correlated, with their well-defined positions on the remotely sensed images. The basic errors during the remotely sensed images' acquisition, that degrade the images' quality and set certain restrictions to the possibilities of the Photointerpretation - Remote Sensing Methodology, are radiometric and geometric ones. So, before the photointerpretation analysis and, more generally, before the remotely sensed images' exploitation, either only radiometric corrections, or only geometric ones or, and most commonly, both radiometric and geometric corrections are required. a. Radiometric Corrections The most common radiometric corrections of remotely sensed images acquired by multi-spectral scanners, are:
the restoration of the 6th line dropout ("loss of data"),
the restoration of the 6th line caused by the abnormality of the signal ("striping"),
the restoration of the horizontal lines parts displacement,
the restoration of the atmospheric diffusion and absorption results, which can eliminate the possibility of data acquisition from the explored region (for some areas of the electromagnetic spectrum). It has to be noted that the images acquired with wavelength λ > 0,7 μm (infrared), are practically clear of the atmospheric diffusion effects.
b. Geometric corrections During the acquisition of digital multi-spectral remotely sensed images, some systematic and non-systematic errors influence the images' geometrical quality. For systematic errors, relevant geometric corrections could be performed depending on:
the movement’s and the orbit’s elements of the carried platform or the remote system,
the element concerning the basic characteristics, the properties, the calibration measurements and the distortions of this particular remote system.
For non-systematic errors, such as:
errors resulting from the accidental altitude divergence of the remote system’s orbit (these errors cause a change of the remotely sensed images scale), and
errors resulting from accidental divergences of the remote system axis, from the specified places of reference (they are, usually, the first perpendicular, the second parallel to the flight axis and the third vertical to the two others previously mentioned),
is required the knowledge of the geodetic co-ordinates of some relating points on the ground, which, simultaneously, are well–defined on the remotely sensed images and consequently their image co-ordinates can be measured in pixels lines and columns. Reference: Rokos, D. “Special Applications of Photointerpretation and Remote Sensing”, NTUA, 1979. Jensen, J. “Introductory digital image processing”, University of South Carolina, 1986.

3.4 Simple Image Enhancement of Remotely Sensed Images	TOP
In order to facilitate the processing of remotely sensed images, we often built image enhancement algorithms. This can be achieved either by:
a. modification of each pixel regardless of the value of the adjacent pixels, or b. modification of each pixel, taking into account the pixels around it.
The most common techniques in image enhancement are:
1. Contrast Enhancement. 2. Pseudo-colour Enhancement - Density Slicing. 3. Image enhancement through basic calculations. 4. Edge enhancement. 5. Filtering. 6. Other transformations such as: Principal Component Analysis, Vegetation Index, Kauth-Tomas/Tasseled Cap transformation, Multiple Discrimination Analysis, Hue-Saturation-Intensity Analysis, Fourier Analysis etc.
Additionally, image reduction and magnification can enhance image resolution and provide more particular information on the whole or a part of the image, as well as a histogram of pixel values along a line on the image.
3.4.1 Contrast Enhancement	TOP
Remote sensing systems record reflected and emitted energy from earth surface materials. Ideally, one material would reflect a tremendous amount of energy in certain wavelengths, while another material would reflect much less energy in the same wavelengths. This would result in contrast between the two types of materials when recorded by a remote sensing system. Unfortunately, different materials often reflect similar amounts of radiant flux throughout the visible and near-infrared portion of the electromagnetic spectrum, resulting in a relatively low contrast image. In an image like this, all characteristics of earth (related to the natural and the socioeconomic reality and their changes through time etc.), become similar to the environment around them and not discernible.
3.4.2 Pseudo-colour Enhancement - Density Slicing	TOP
Pseudo-colour enhancements of an image intend to convert it from greyscale to colour (not necessarely to natural colours or like false infrared), in order to use all values of the electromagnetic spectrum. The discrimination ability of the human eye is usually limited to 16 different grey levels. This limitation may be overcome if greyscale images are converted to pseudo-colour. This conversion allows the utilisation of the entire range of the image brightness values or portions of it in different hue values. Density Slicing is the technique that assigns different hue values to each greylevel, in order to make the discrimination of different homogenous zones easier. Usually, the selection of the successive portions is arbitrary and this causes loss of information. Studying the histogram and selecting the portion that refers to discernible homogenous zones could prevent this. In this case the final image after density slicing control will be enhanced and much easier to interpret.
3.4.3 Image enhancement through basic numerical calculations	TOP
A remotely sensed image of the earth is a 3-D numerical matrix.
X, Y axes constitute a Cartesian reference system of spatial features.
Z-axis refers to the pixels' digital values for the amount of the energy emitted/reflected from every snapshot of earth surface.
The pixel’s digital value range is determined by the radiometric resolution of the sensor: 0 – 63 for radiometric resolution of the sensor 6 bits (IRS) 0 – 127 for radiometric resolution of the sensor 7 bits 0 – 255 for radiometric resolution of the sensor 8 bits (Landsat TM) 0 – 511 for radiometric resolution of the sensor 9 bits 0 – 1023 for radiometric resolution of the sensor 10 bits (NOAA) 0 – 2047 for radiometric resolution of the sensor 11 bits (Ikonos) The amount of the radiation emitted/reflected from specific features on the earth surface depends not only on their physical, chemical or biological characteristics, but also on a series of other factors (terrain features, sun ratioing etc.). This could cause difficulties during image interpretation, which may be eliminated by applying basic numerical calculation (subtraction, rationing) on the proper components (bands) of a multi-spectral digital remotely sensed image. So, we may have image enhancement with the following characteristics: a. Addition of brightness values We can make an enhanced image by adding two or more remotely sensed images of the same date, time, region and dividing the sum with the number of the remotely sensed images. This image is the average of its components and is released from the total noise allocation. b. Subtraction of brightness values By subtracting the appropriate two bands of two multi-spectral satellite images of the same area acquired at the same date in different years or in different seasons of the same year, we can view the annual/seasonal changes in the area, because of the nullification of the same pixel’s digital values. c. Ratioing of Remotely Sensed Images The image interpretation potential of a multi-spectral digital remotely sensed image is often bounded by the fact that specific features of one homogeneous zone reflect/emit electromagnetic radiation in different intensity. So, the features appear on the image in different grey tones. The terrain features cause this – humidity, seasonal change etc. Ratioing the appropriate two components (bands) of the remotely sensed images, we can reduce the impact of the above mentioned factors.
3.4.4 Edge enhancement	TOP
Image interpretation of a remotely sensed image may be easier if the edges of the objects/characteristics are enhanced by an edge enhancement operation, so that their shapes and details are enhanced. Generally, what eyes see as pictorial edges are simply sharp defferences in brightness value between two pixels. The edge enhancement of a remotely sensed image can be performed, either with linear edge enhancement or with non-linear edge enhancement.
3.4.5 Filtering	TOP
The filtering of a remotely sensed image aims at the enhancement/improvement of the image, either with the elimination or compression of certain spatial frequencies and linear characteristics which abstract to the interpretation of other interesting characteristics (road network etc.) or with the enhancement of spatial frequencies and linear characteristics which concern us the most (boundaries of water resources). The procedure, which is used to decompose a remotely sensed imagery to its different spatial frequency components, is Fourier Analysis. Applying Fourier Transform on the original remotely sensed image is a way to enhance certain interesting spatial frequency components for image interpretation, contrary to others that do not interest us. The algorithms, used for such an enhancement/improvement, are filters such as:
low pass filters
high pass filters.
High pass filters emphasize high frequency spatial characteristics, thus enhancing the linear characteristics of a remotely sensed image. Low pass filters, on the other hand, emphasise low frequency spatial characteristics, causing a “smoothing” of the remotely sensed image.
3.4.6 Special enhancement transformations	TOP
Some of the special enhancement transformations used for natural resource inventories and monitoring, are related to principles, methods and techniques already mentioned. So, we can use transformations, such as:
Principal Component Analysis of a remotely sensed image
Multiple Discriminant Analysis of a remotely sensed image.
Reference: Rokos, D. “Special Applications of Photointerpretation and Remote Sensing”, NTUA, 1979. Jensen, J. “Introductory digital image processing”, University of South Carolina, 1986.
3.5 Classification	TOP
Remote sensing data are records of reflected and emitted electromagnetic energy presented as picture like images. In order to extract meaningful information out of the data, an interpretation of the image has to be carried out. The image interpretation is made with the help of an electronic computer, where use of mathematical algorithms takes place, provided of course, that the data are in digital form. The classification methods of the computer-assisted approach are:
Supervised classification: Representative sample sites of known land cover type (from a ground truth survey or from a map e.g. olives, forest, sea), called “training areas”, are selected in order to compile a numerical “interpretation key” that describes the spectral attributes of each feature type of interest. In other words these pixels from the “training areas” will be used to “train” the computer so that each pixel from the data set will be compared numerically to each category in the interpretation key and labeled with the category it “looks most like”. There are a number of numerical strategies (means, standard deviations etc.) that can be employed to make this comparison between unknown pixels and pixels derived from the “training areas”. The comparison is always made with the use of a specific classification algorithm, like the Gaussian Maximum Likelihood, the Parallelepiped and the Minimum distance to means algorithms.
Unsupervised classification: In this approach no training samples are used. The pixels from the remote sensing data are classified by the method of cluster analysis, which can identify natural groupings of patterns. The classes that result from unsupervised classification are spectral classes. Because they are based solely on the natural groupings in the image values, the identity of the spectral classes will not be initially known. The nature of each grouping is determined afterwards by ground truth surveys.
Unsupervised classification is not generally as effective as supervised classification because of the absence of training sets to control the results, especially when classes are only marginally separable. However, supervised classification is still too slow to handle a massive influx of satellite multi-spectral data. The results of the classification from the computer-assisted approach can be presented as a thematic map, which can be stored in digital form. The computer can also display numerical information on the area of the mapped classes or the frequency of the occurrence of each class and other useful statistical data if required. Reference: C.P.Lo., “Applied remote sensing”, University of Georgia. Jensen, J. “Introductory digital image processing”, University of South Carolina, 1986.