80
2: Raghuveer M. Rao, Manoj K. Arora
In post classification comparison, the two images are first classified using
any of the techniques mentioned in Sect. 2.6. These two classified images are
then compared pixel by pixel to produce a change image representing pixels
placed in different classes. For this method to be successful, the classification
of images should be performed very accurately.
Change detection has also been performed using many other algorithms
such as change vector analysis (Lambin and Strahler 1994), principal component analysis (Mas 1999) and neural networks (Liu and Lathrop 2002). A recently developed MRF model based change detection algorithm is presented
in Chap. 12.
2.8
Image Fusion
During the last several years, enormous amount of data from a number of
remote sensing sensors and geo-spatial databases has become available to the
user community. Often, the information provided by each individual sensor
may be incomplete, inconsistent and imprecise for a given application (Simone et al. 2002). The additional sources of data may provide complementary
information to remote sensing data analysis. Therefore, fusion of different
information may result in a better understanding of the environment. This is
primarily due to the fact that the merits of each source may be tapped so as
to produce a better quality data product. For example, a multispectral sensor
image (at fine spectral resolution) may be fused with a panchromatic sensor
image (at high spatial resolution) to generate a product with enhanced quality,
as spectral characteristics of one image are merged with spatial characteristics
of another image to increase the accuracy of detecting certain objects, which
otherwise may not be possible from individual sensor data. Thus, image fusion
may be defined as the process of merging data from multiple sources to achieve
refined information. The fusion of images may be performed to (Gupta 2003):
1. sharpen the images
2. improve geometric quality of images
3. improve classification accuracy
4. enhance certain features
5. detect changes
6. substitute missing information in one image with information from another image
The complementary information about the same scene may be available in
different forms and therefore different kinds of fusion may be adopted. For
example, data from different sensors (Multi-sensor fusion), at different times
(multi-temporal fusion), different spectral bands (multi-frequency fusion),
2: Raghuveer M. Rao, Manoj K. Arora
In post classification comparison, the two images are first classified using
any of the techniques mentioned in Sect. 2.6. These two classified images are
then compared pixel by pixel to produce a change image representing pixels
placed in different classes. For this method to be successful, the classification
of images should be performed very accurately.
Change detection has also been performed using many other algorithms
such as change vector analysis (Lambin and Strahler 1994), principal component analysis (Mas 1999) and neural networks (Liu and Lathrop 2002). A recently developed MRF model based change detection algorithm is presented
in Chap. 12.
2.8
Image Fusion
During the last several years, enormous amount of data from a number of
remote sensing sensors and geo-spatial databases has become available to the
user community. Often, the information provided by each individual sensor
may be incomplete, inconsistent and imprecise for a given application (Simone et al. 2002). The additional sources of data may provide complementary
information to remote sensing data analysis. Therefore, fusion of different
information may result in a better understanding of the environment. This is
primarily due to the fact that the merits of each source may be tapped so as
to produce a better quality data product. For example, a multispectral sensor
image (at fine spectral resolution) may be fused with a panchromatic sensor
image (at high spatial resolution) to generate a product with enhanced quality,
as spectral characteristics of one image are merged with spatial characteristics
of another image to increase the accuracy of detecting certain objects, which
otherwise may not be possible from individual sensor data. Thus, image fusion
may be defined as the process of merging data from multiple sources to achieve
refined information. The fusion of images may be performed to (Gupta 2003):
1. sharpen the images
2. improve geometric quality of images
3. improve classification accuracy
4. enhance certain features
5. detect changes
6. substitute missing information in one image with information from another image
The complementary information about the same scene may be available in
different forms and therefore different kinds of fusion may be adopted. For
example, data from different sensors (Multi-sensor fusion), at different times
(multi-temporal fusion), different spectral bands (multi-frequency fusion),