Apply Colour Sequences to Enhance Filter Results Operations What Do I Need? Filter Single band images from the SPOT and Landsat platforms can sometimes appear flat (i.e., they are low contrast images). These images can be enhanced in many ways, including automatic and manual contrast stretching and image filtering. Contrast stretching enhances the image without altering the underlying data values; while image filtering generates a new image with sharpened edges but altered underlying data values. The contrast stretch is the most common way to enhance low contrast images. The technique can be used to emphasize medium and low frequency features in the images or highlight dark or bright areas. It is common to perform multiple contrast enhancements on a single images. Manipulating the colour and/or contrast of an image is often sufficient to interpret an image without applying filters. Filters enhance images by detecting and emphasizing high or low frequency changes in values across the image. The MFworks Filter operation can be used to apply a smoothing filter and several different edge detection/enhancement filters. There are dozens of enhancements that can be performed on remote sensing imagery. For example, it is common to merge individual bands of an image to generate a false colour composite. The MFworks Merge operation can be used for this purpose. The Merge operation can also be used to add single colours to greyscale images. You have probably seen images of the surfaces of Mars coloured orange or the surface of Venus coloured gold. This is done to enhance the feature perception ability of the viewer. You can also add or multiply images by thematic map layers with the same coverage to add context to the thematic map layers. For example, you could add satellite imagery to geologic or soil map layers to give context to the geologic or pedologic units. HD-CSF-1
Example Three Enhancements for a NIR (Near Infrared) SPOT Image The map layer named XS3 (NIR) is a portion of a SPOT satellite near infrared image of the City of London, Ontario, Canada. Near infrared imagery displays the quantity of infrared wavelengths being radiated or reflected from the various surfaces in the image. Different types of surfaces radiate/reflect infrared energy at different intensities. The more energy received by the sensor, the brighter the image pixel will be. The map layer named XS3 (NIR) is fairly low contrast. The most striking features are the black bodies which are ponds and the very bright patches which represent healthy vegetation: This image can be improved by applying a different colour sequence to the legend. With the Legend window as the active window, Select All is chosen from the Edit menu, then Colour Sequence is selected from the Legend menu. This opens the Sequencer dialog box. The factory default black to white, sequential colour sequence was applied automatically when the image was imported. Changing the colour distribution to Change in Zone Area will improve the image contrast: HD-CSF-2
As you can see, this change greatly improves the contrast of the image. However, due to the nature of near infrared energy, the features are not very sharp. This image can be improved even more by applying an edge enhancement filter: You may have to experiment with the different filters to see which one yields the most desirable results. After some experimentation, the Filter (Laplacian) operation is chosen. The Filter operation is selected from the Operation menu. This causes the Filter dialog box to be displayed. The map layer named XS3 (NIR) is specified in the Map drop-down list and the Laplacian Type 1 filter radio button is clicked: If you were performing this operation from the Script window the statement would be: HD-CSF-3
XS3 (NIR) Lapl = Filter XS3 (NIR) Laplacian1; The Laplacian 1 filter sharpens the edges of the fields, roads, ponds, and buildings in the image. The Change by Zone Area colour sequence, as above, has been applied to the resulting map layer to enhance image contrast. Note: when you apply a filter to a map layer, the original zone values are lost: Sometimes a single pass of a filter is not sufficient to enhance or sharpen an image for certain applications. You can apply the same filter more than once. Here is the result of applying the Filter (Laplacian 1) operation twice to the map layer named XS3 (NIR). As you can see, the fields, roads, and HD-CSF-4
neighbourhoods are more sharply defined in this map layer than in the single filtered map layer: A third way to enhance a remote sensing image is to apply a colour sequence other than black to white to the zone values. Colour sequences can be used to visually group value ranges together. For example, you could apply a red to purple colour sequence to the filtered map layer created above. With the Legend window as the active window, Select All is chosen from the Edit menu, then Colour Sequence is selected from the Legend menu. This opens the Sequencer dialog box. A red to purple colour sequence is specified with a counter-clockwise path though colour space. A colour distribution of Change in Zone Area is used to improve the image contrast: As you can see, this colour sequence greatly enhances the interpretability of this image. Areas that were dark in the original image now stand out as bright red. Areas that were bright in the original image are all coloured HD-CSF-5
purple. Any number of colour sequences can be developed for your remote sensing images using the MFworks colour facilities: Example Three Enhancements for a Red SPOT Image The map layer named XS2+P (Red) is an image depicting the reflectance of red light from a portion of a SPOT satellite image of the City of London, Ontario, Canada. This image displays the quantity of red wavelengths being reflected from the various surfaces in the image. Different types of surfaces reflect red wavelengths at different intensities. The more energy received by the sensor, the brighter the image pixel will be. The map layer named XS2+P (Red) is fairly low contrast. Most features, apart from some of the larger buildings, are a uniform dull grey: HD-CSF-6
The contrast can be improved by changing the colour sequence from the default black to white, sequential distribution that was applied automatically when the image was imported to a black to white colour sequence with colours distributed by Change in Zone Area. With the Legend window as the active window, Select All is chosen from the Edit menu, then Colour Sequence is selected from the Legend menu. This opens the Sequencer dialog box. The Change in Zone Area distribution is chosen to improve the image contrast: As you can see, features now stand out starkly. This is a very high contrast image. While features can be easily discerned, it may appear too contrasty for certain applications. Also, the details of some features are not as sharp as they can possibly be. This image can be improved even more by applying an edge enhancement filter: You may have to experiment with the different filters to see which one yields the most desirable results. After some experimentation, the Filter (Laplacian 1) operation is chosen. The Filter operation is selected from the Operation menu. This causes the Filter dialog box to be displayed. The map layer named XS3 (NIR) is specified in the Map drop-down list and the HD-CSF-7
Laplacian Type 1 filter radio button is clicked: If you were performing this operation from the Script window the statement would be: XS2+P (Red) Lapl = Filter XS2+P (Red) Laplacian1; The Laplacian 1 filter sharpens the edges of the fields, roads, ponds, and buildings in the image. Note: when you apply a filter to a map layer the original zone values are lost. The default black to white, sequential distribution colour sequence was applied to this map layer when it was created, yielding a flat (low contrast) image: HD-CSF-8
The contrast can be improved by changing the colour sequence to a black to white colour sequence with colours distributed by Change in Zone Area. With the Legend window as the active window, Select All is chosen from the Edit menu, then Colour Sequence is selected from the Legend menu. This opens the Sequencer dialog box. The Change in Zone Area distribution is chosen to improve the image contrast: The combination of edge detection filtering and contrast enhancement has greatly improved the interpretability of this image. All the features of the landscape that are discernable at this cell resolution are well defined: The Change in Zone Area colour sequence does not always produce such satisfactory results. Sometimes the image can appear too contrasty. Also, during image interpretation you may wish to emphasize certain tonal ranges over others. If such is the case, then you will have to apply a manual colour stretch. HD-CSF-9
Using the Histogram Technique to Enhance a Filtered Image To apply a manual colour enhancement to a map layer, click on the Legend window for the map layer to make it active, then choose Format from the Legend menu. The Legend Format dialog box will open: Legend Entry Prototype Area Legend Element Holding Area Drag the Value, Area, and Text elements from the legend entry prototype area to the legend element holding area. Move the Colour element to the right in the legend entry prototype area. Place the Value element back into the legend entry prototype area, this time to the left of the Colour element: Click on the Colour element to make it active then drag the grow handle to the right until the Colour element fills the legend entry prototype area. HD-CSF-10
Finally, select the Bar Format option and set Max to 1%. Finally, click on the Apply button: The Legend window is now formatted to appear as a vertical bar graph with lowest values at the top and highest at the bottom. Scroll through the list of values and you will see the Colour element gradually get longer and then shorter. Most of the values fall in the mid-range and are assigned middle grey values. Scroll back to the top of the window and select the first legend entry. Scroll down until the Colour elements start to get longer. Hold down the Shift key and select one of the legend entries. This will cause all of the entries from your first selection to your second selection to become selected: HD-CSF-11
With these entries selected, choose Group Selection from the Legend menu. All the selected legend entries will be grouped into a single legend entry. Assign the colour black to this group: Now scroll to the bottom of the Legend window and select the last entry. Scroll upward until you see the Colour elements start to get longer. Hold down the Shift key and select one of the legend entries. This will cause all the entries from your first selection to your second selection to become selected: HD-CSF-12
With these entries selected, choose Group Selection from the Legend menu. All the selected legend entries will be grouped into a single legend entry. Assign the colour white to this group: Grouping the upper and lower outlier values is equivalent to trrimming values beyond 1 to 2 Standard Deviations. If you have a multi-modal distribution you can further enhance the image by applying separate colour sequences to dominant classes. To complete the contrast stretch, choose Select All from the Edit menu then choose Colour Sequence from the Legend menu. Specify a black to white, linear sequence with a Sequential distribution. Click on the Apply button: This method allows you to set the contrast of the image exactly the way that you want it. You can make the contrast higher by including more high and low values in the uni-colour groups, or you can make the contrast less HD-CSF-13
extreme by including fewer values in the high and low uni-colour groups. Here are the results of the above contrast stretch: HD-CSF-14