Clicking on the 'Display' button on the pink canvas will give you the option 'New frame display'.
From the display options, 'Single Raster' produces a grey scale image based on just one of the data layers.
The default grey scaling for any layer selected will initially be between the
minimum and maximum possible for that layer (e.g. 0 and 255 for a spectral
layer). This can produce an image with low contrast. The 'Colours' menu
option 'Set grey range to min/max' allows linear rescaling between the
values actually occurring in this layer, for better contrast.
'Mask layer' allows you to look at the boolean mask layer, if there is one in the dataset, as though it were a black&white image.
'RGB Display' prompts for the layers required, and produces an RGB image. The checkboxes allow details of each of the 3 layers to be viewed.
Below you can see:
3 of the Landsat bands in the example dataset: these stretch over the whole area of the map
3 of the verification layers from the example dataset: here, only the central area has been surveyed, and some pixel mixing can be seen as well as some homogeneous areas.
3 of the fuzzy membership layers from the example dataset. These have been generated by unsupervised classification from the Landsat data, so they cover the whole extent.
'Subtracted' allows one layer to be subtracted from another - e.g. a predicted membership layer and the verification data for the same class. This means you don't have to store the subtracted layer in the dataset but can view it on the fly.
The colour scale shows blue for negative values, red for positive.
NB: the subtraction module is in the process of being changed, and currently shows a colour range from GREEN to RED. The documentation will shortly be updated. (2/8/04)
In all these static displays, there is the capacity to query the data, and to
select particular pixels and look at their neighbourhoods. These functions are
carried out using the 'Mouse' menu on the left.
'Query raster' gives details of any particular pixel on which you click. When switched back off, this box with details will disappear. While this box is open, it will respond to any single click within the image by giving details of that particular pixel.
NB: The layer name is currently displayed as 'Blank' (02/08/04)
Selecting single pixels.
'Select a single pixel' allows a single pixel to be selected and activated. This is important in linked visualisations, where this particular pixel can be highlighted in any graphs & charts which are open. The pixel is currently highlighted in red. The 'select single pixel' option stays switched on until the user switches it off.
Selecting groups of pixels.
The frame also allows a group of pixels to be selected, by clicking, drawing a box or digitising a polygon. All these options are available on the 'Mouse' menu.
'Select group (point)' records and highlights a sequence of pixels selected with the mouse, ending with a double-click. The command is then switched off again.
'Select group (box)' allows a box of pixels to be selected, by pressing the mouse on one cormer, dragging it across to the opposite corner and releasing it. The box shows up when coordinate selection is complete.
'Select group (digitise)' allows a polygon to be defined and highlighted, by a sequence of clicks ending with a double click.
Whenever a group of pixels is selected in this way, from either a static image or an animation, their identities are sent to all the other open visualisations and graphs, where they are also highlighted.
'Get cross section of memberships' allows a pixel to be selected, and a 4-way histogram of the fuzzy memberships in its neighbourhood displayed. To test this, check the option, and then click on any pixel in the image. Then specify a window size. The layers defined as 'fuzzy memberships' within the raster set are automatically queried to produce cross sections of the area around this pixel.
The 'C' represents the central pixel, and the 4 buttons at the top can be used to display the horizontal section
( - ), the vertical ( | ), the upslant left to right ( / ) and the downslant ( \ ).
To close a display frame, select Clear>Close Frame from the menu at the top.
If you wish to get a quick view of the values in other data layers, while you are querying a raster image, you can set up 'gauges' which relate to those layers, as follows;
From the 'Gauges' menu on the display frame, select 'Add a gauge', and select the raster layer which you wish to monitor.
While gauges are active, any click on the existing image will show the
values in the relevant layers, at that point. If the point selected is covered
by the mask layer, then the needles on the gauges disappear. 'Remove all
gauges' will close them all down.
Open a new display frame, and view the spectral data in the 'Peak' test file.
Rescaling from minimum to maximum (a contrast stretch) helps to pick out the
Any 3 bands of the same type (e.g. spectral, fuzzy memberships), can be viewed as an RGB image.
View the entropy layers produced in the last page, and the subtracted product. To see the origin of the entropy statistic, view the entropy layer derived from the fuzzy membership layers, and select the 'Get cross section of memberships' option. Select a pixel in an area of high entropy, with a window of about 11 pixels. The pixels should all be fairly mixed, so the columns of the cross-section graph will contain numerous stripes of different colours. In an area of low entropy, the pixels will be more dominated by single classes.
One thing which is of interest is the extent to which the verification data
and the fuzzy memberships differ, for each class. For example, do any of the
fuzzy membership layers seem to correspond to the verification landcover
classes? Producing added and subtracted images could help to verify whether this
is the case.
NEXT - Animations