successful demonstration for a specific image in the scientific literature may be the
end result of a long process of methodological development tested on that specific
image, but not presented as part of the published paper. Therefore the potential
transferability of methods to other sites must always be assessed critically.
4.2.1 Data and Processing Requirements
Subsequent sections give an overview and examples of different image processing
techniques for mapping coral reef biophysical parameters. Emphasis will be given
not only to the type of data that can be extracted, but also to what is required in a
practical sense to apply each method. In general, factors to consider when evaluating the potential of published methods are:
Is the method appropriate for the imagery that is available?
• Does the method rely on hyperspectral imagery? What are the number of bands,
spectral width of bands and radiometric accuracy of instruments used in successful demonstrations?
• Is the spatial resolution of the imagery appropriate? What is the scale of benthic
heterogeneity at the study site and can the method handle this level of withinpixel mixing?
• If the imagery is already acquired, is it of good quality? An image with substantial sunglint or atmospheric haze may not give as good results as the
imagery used in a published paper. Again, there is likely to be a bias as the best
published demonstrations will likely have used the highest quality images (see
Brando et al. 2009 as a rare exception).
• If the imagery is not already acquired, what is the risk that it will be of insufficient quality to support the analysis? For example, commercial satellite data
providers do not always provide quality assurance for sunglint. Is an airborne
campaign subject to time, personnel or cost constraints that may compromise the
quality of data?
What are the processing requirements and how will these be achieved?
• Is atmospheric correction required? If so, which method or software should be
used and what data will be required (see Sect. 4.2.3)?
• Is an air–water interface correction required, either for sunglint removal or to
align the magnitude or type of reflectance with modeled or in situ data (e.g.,
above-surface or below-surface reflectance)?
• What level of geo-correction is required to align with ground truth data? Is
alignment even feasible given the spatial scale of the image and available or
potential ground truth data?
• Is the analysis procedure a routine that can be found in standard image processing software, or does it require custom code?
4 Hyperspectral Applications
87
end result of a long process of methodological development tested on that specific
image, but not presented as part of the published paper. Therefore the potential
transferability of methods to other sites must always be assessed critically.
4.2.1 Data and Processing Requirements
Subsequent sections give an overview and examples of different image processing
techniques for mapping coral reef biophysical parameters. Emphasis will be given
not only to the type of data that can be extracted, but also to what is required in a
practical sense to apply each method. In general, factors to consider when evaluating the potential of published methods are:
Is the method appropriate for the imagery that is available?
• Does the method rely on hyperspectral imagery? What are the number of bands,
spectral width of bands and radiometric accuracy of instruments used in successful demonstrations?
• Is the spatial resolution of the imagery appropriate? What is the scale of benthic
heterogeneity at the study site and can the method handle this level of withinpixel mixing?
• If the imagery is already acquired, is it of good quality? An image with substantial sunglint or atmospheric haze may not give as good results as the
imagery used in a published paper. Again, there is likely to be a bias as the best
published demonstrations will likely have used the highest quality images (see
Brando et al. 2009 as a rare exception).
• If the imagery is not already acquired, what is the risk that it will be of insufficient quality to support the analysis? For example, commercial satellite data
providers do not always provide quality assurance for sunglint. Is an airborne
campaign subject to time, personnel or cost constraints that may compromise the
quality of data?
What are the processing requirements and how will these be achieved?
• Is atmospheric correction required? If so, which method or software should be
used and what data will be required (see Sect. 4.2.3)?
• Is an air–water interface correction required, either for sunglint removal or to
align the magnitude or type of reflectance with modeled or in situ data (e.g.,
above-surface or below-surface reflectance)?
• What level of geo-correction is required to align with ground truth data? Is
alignment even feasible given the spatial scale of the image and available or
potential ground truth data?
• Is the analysis procedure a routine that can be found in standard image processing software, or does it require custom code?
4 Hyperspectral Applications
87