Image Processing
With PEP and a lot of NASA data you can explore another planet. But sometimes, that data is hard to interpret. Maybe the data is too low in contrast, maybe it has too much noise, maybe you need to blend different data sets together. To solve these kinds of problems you need to adjust or tweek the data. So, PEP includes a suite of easy to use, interactive image processing filters you can apply to any NASA data set. The PEP image processing filters mostly affect the data displayed in the map view, not the 3D view.
Each NASA data set (for example, Viking global imagery, MOLA elevation, THEMIS albedo, etc) is displayed in a different layer. These layers can be individually turned on or and reordered so that layer can be drawn above or below any other layer. Each layer also has a its own set of image processing filters. These filters are applied in order to convert the raw NASA data into the image you see on the map.
To access the image processing controls, click the "Tools" menu item and select "Display Image Processing Controls". A new window will appear:
From this single window you control all the image processing filters for all the layers. The first step is to select the layer you want to operate on. This is done by clicking on "Select Layer" and choosing the right layer from a pop-up list. After you have selected a layer, the title of the window is changed to match the layer and its image processing filters are displayed in the bottom of the window.
Often, each layer is initialized with multiple image processing filters. If you want to adjust one of these existing filters (say, to adjust contrast or change colors) you select the filter by clicking on the filters tab. A list of all the current filters for the selected layer will pop up and you simply select the one you want.
Nearly all filters contain some controls that you can adjust. For example, the above screen shot shows the user interface for the False Color filter. By default the Global Imagery layer is colored gray and opaque. If you want to see something more exotic, you can change the color to red.
Below the controls for each filter are two buttons, labeled "Disable" and "Delete". These buttons appear on the controls for each and every filter. When you select "Disable" the filter is turned off so you can see what effect it is having on the data. Naturally selecting "Delete" will delete the filter.
Besides adjusting existing image processing filters, you can create new ones. Click on "Create New Filter" to see the list of all possible filters and select the one you want. Its controls will automatically appear.
PEP includes the following filters: brightness and contrast, false color and transparency, sharpen, blur, threshold, edge detection, general convolution, gradient and median.
The Image Processing Filters
Brightness And Contrast Filter
Planetary probes gather data for an entire planet. Its camera must be prepared to take pictures of the darkest and brightest regions. Sometimes, when you're looking at data from one specific region, the image can be too dark, too bright or completely lacking in contrast. In this case, you need to adjust the layer's Brightness And Contrast Filter. With this filter you can fiddle with the map image until the details you are trying to see are more obvious. Simply move the sliders to change the image.
By default, the Global Imagery layer creates a Brightness And Contrast Filter to increase contrast by 400%. To see what the original Viking images look like move the contrast slider down to 0. On the map you'll get a much more muddled image where all fine detail disappears.
Planetary probes generally don't send back color photographs. Instead the sensors generate some binary information that can be displayed in a wide variety of ways. It isn't hard to take this data and display it as picture where everything some shade of gray or yellow or blue. With the "False Color" filter, you just click on the color you want to use.
By adjusting the "Transparency" slider you control how much of the underlying layers you can see. To continue the above example, you can display the thermal inertia data subtly showing just a hint of yellow with the underlying imagery very visible or you can move the slider so the yellow obliterates what is under it. The former might be good when you are exploring a region while the latter is useful if you want to capture an image for a presentation.
Sharpen Filter
The sharpen filter helps highlight fine detail. Just move the slider to control how much the image is sharpened. Typically, since the images are a little noisy this filter isn't very useful.
Sharpen works by applying a 3x3 Laplacian convolution array to the image. As you vary the amount of sharpening, the parameters in this convolution array change.
Blur Filter
The blur filter helps clean up noisy images. Simply move the slider to adjust the amount of blur applied.
Generally, all the data (from the lowest values to the highest values) for a layer is displayed. To reduce clutter, you might just want to just display pixels with a high data value and have the low values disappear. You can do this with the Threshold Filter. By moving the slider you specify the minimum value a pixel has to have to be displayed. This is useful after an Edge Detection filter to remove the weakest edges.
Note that the threshold slider values range from 0 to 255. That is, they are scaled data values. If you're dealing with, for example, the elevation layer you might want to set a threshold at a specific number of meters. Sadly, you can't do that. Instead you must currently use the dimensionless scaled image values. If that is a problem please file a complaint!
Edge Detection (Sobel Gradient) Filter
This filter will run an edge detection algorithm over the layer's data. Sometimes I'll apply an edge detection algorithm to the Martian elevation data and display it over top of the global imagery. Where ever there is a sharp change in elevation, a line will be drawn. This makes it easier to see where there might be something interesting going on. With Venus data I've applied an edge detection algorithm to the imagery data to make it easier to find volcanoes. With the edge detection filter, the caldera on top of Venusian volcanoes stand out nicely.
Often before you apply the edge detection filter you should reduce the amount of noise in the image. Otherwise, each speck of noise will appear as an edge. To help eliminate this "speckle noise" create a Blur Filter before the Edge Detection filter.
I should note that all image processing functions are computed individually on each NASA data file. This can sometimes result in the the edges of each file becoming very prominent. Since the elevation data is store in a single file, this isn't an issue for it. However, if you apply the Edge Detection filter to the Global Imagery the boundaries of each NASA file will show up as a very hard edge.
To quote Sun's Java Advanced Imaging documentation: the "GradientMagnitude" operation is an edge detector which computes the magnitude of the image gradient vector in two orthogonal directions. PEP makes use of the Sobel operators.
General Convolution Filter
If you know something about image processing and kernel math, there might be some specific image processing filter you need. To make it easy for you to do whatever you want, PEP includes a General Convolution Filter. Its controls are a kernel array that will be applied to the data. You can even specify the size of the square kernel you want, up to an 11x11.
Gradient Filter
Gradient filters combine the results of two different convolutions to create an image that can highlight edges along a particular direction. To use this filter you really know to know something about kernel math, convolutions and linear algebra. But, if you think I was vastly, hugely wrong for picking the loser Sobel operator instead of the cool Prewitt, here is where you can set the world right.
Median Filter
The median filter might be identical to the blur filter. After I have a chance to look into it more deeply, it may disappear. Until then, you can use it to reduce noise. It computes the value of each pixel by computing the average of those pixels around it. You specify the size of the averaging area. A larger averaging area will blur the data more.
Previous Home Next