In this topic you have explored concepts, examined algorithms and developed
skills of processing EO data. Your studies have focused on information
contained in the spatial domain. The spatial domain is very important
in Earth observation by remote sensing. The most significant element is
that of spatial scale or resolution as this effectively controls the information
content present within a remotely sensed image.
Remote Sensing and Spatial Resolution
The spatial resolution of remotely sensed imagery is a strong controlling
factor in determining which objects can be distinguished within a scene.
Spatial resolution can pose problems in operational remote sensing associated
with the ‘mixed pixel’ problem, where pixels are characterised by heterogeneity
rather than purity. Image variance is directly related to the size of
target objects present within the image scene and the spatial resolution
of the imaging sensor so the choice of imagery will be application dependent.
As a budding image analyst you should be able to select the most appropriate
sensor type and spatial resolution for a particular remote sensing application.
The spatial structures present within an image scene may be preferentially
enhanced by a range of algorithms. The most widely used technique is the
spatial frequency filter which allows the analyst to emphasise or suppress
different spatial frequency components including large homogenous area,
edges and directional features. An alternative technique, the Fourier
Transform allows image analysts to process images using an alternative
co-ordinate space known as the power spectrum, which highlights spatial
frequencies and orientations present within the original image scene.
This technique can be used to remove spatial noise present within an image
scene or alternatively as an improved (though computationally demanding)
filtering method. The variogram is a measure of spatial dependence variance
within a remotely sensed image scene. It can be used in image interpretation
and allows the image analyst to determine the presence, and distribution,
of any spatial structures or patterns within the scene.
The analysis of the spatial components of remotely sensed imagery can
be performed by visual interpretation or by automated techniques using
texture and contextual information. Spatial properties such as texture,
scale, pattern and context may be employed in the analysis and interpretation
of remotely sensed imagery. The use of regions or ‘objects’ within a scene
allows the image analyst to combine spatial and spectral information in
the interpretation and classification of images.
Multisource Remote Sensing
The processes (and the scales of these processes) operative within the
natural and human landscape are many and varied. Often these processes
cannot be identified or analysed using single source imagery. Multisource
remote sensing, often known as fusion techniques, allow the analyst to
combine the spatial and spectral properties of different images and allow
the discrimination of finer target objects from the image scene.