Enhancing (Filtering) an Image
Image enhancement differs from image processing in that it allows you to operate directly on the image, including using
filters that perform neighborhood operations on groups of pixels to help enhance
edges, reduce noise and grain, etc. The operations available were chosen because
of their usefulness in enhancing the types of images you’ll most often encounter. Some of the names may be unfamiliar because they are
more typically used in the graphics fields. The best way to become familiar
with their usefulness is experimentation. You can often get a very good affect
by combining two or more filters, as long as they don’t completely cancel each other out.
You can add any number of enhancements to an image. To undo the enhancements,
you can either use the Revert command, or make any adjustment to the image processing (this re-creates the image).
To apply an image enhancement, select the image, then click Enhancement/Filtering from the Image menu. This will display the Image Enhancement and Filtering dialog. The
individual enhancements/filters are described below.
Generic Filters
There are a number of filters available that perform operations on either a
3x3 or 5x5 neighborhood of pixels. These filters transform a pixel's value based
on the surrounding pixels, typically providing a means of smoothing the image
for noise reduction. Some, however, are useful for sharpening and edge
detection.
Most (but not all) of the filters come in either 3x3 or 5x5 versions. This
refers to the size of the pixel matrix used to filter the image. In most cases,
you would choose a 3x3 for lower resolution images, and a 5x5 for higher
resolution. In some cases, best results may depend on the magnification and overall “grain” size of image elements.
In all cases, first select the filter, then hit the Apply button for apply the filter to the image.
Smoothing Filters
Average and Gaussian
Average and Gaussian filters come in both 3x3 and 5x5. They are both
smoothing filters, but produce slightly different results depending on the noise or
graininess of the image.
Low Pass
The 3x3 and 5x5 low-pass filters provide a simple smoothing function, usually
resulting in a slightly blurred image.
Median
This filter (3x3 only) provides a very fine grain/noise reduction.
Sharpening Filters
High Pass
As the name implies, both 3x3 and 5x5 high-pass filters provide the opposite
effect of smoothing filters –they enhance the grain of the image, thereby sharpening it. High-pass filters
by themselves may provide too much of a grain effect, however they can
sometimes provide a better effect when used in conjunction with average and gaussian
filters.
Edge Detection
LaPlacian
These filters (LaPlacian 3x3 and LaPlacian Modified 3x3) provide simple
edge-detection.
Prewitt Compass
The prewitt compass filters (3x3 only) are directional edge detection filters.
These filters are especially useful for enhancing images with topographic
features (for example, SED images). The direction specified (N = North, E = East,
etc.) determines the edge that will be enhanced. The selected edge will
appear much brighter then the rest of the image. These filters can be very useful
when overlaying colored elemental images on an SED.
Prewitt Gradient
Both the X and Y Prewwit Gradient filters are direction edge detection filters
similar to the Prewitt Compass filters, but are designed specifically for
horizontal (X) and vertical (Y) edge detection.
Sharpening
There are two special filters provided specifically for sharpening an image:
Standard Sharpening
The standard algorithm applies a high-pass filter to the image, and provides
an adjustment to the degree of sharpness. Adjust the Sharp Slider between 0 (less sharp) and 25 (more sharp), then hit the Apply button to apply the filter. This filter provides a high degree of
sharpening, but it will often create some extra graininess.
UnSharp Mask
This filter (with the counterintuitive name) is actually a sharpening filter.
However, it is a two step filter – the first step sharpens the image, the second step uses data from the
original image to reduce the grain effect. In this case, the weighting factor
determines the relative importance of the original image data in producing the output.
Adjust the Weight Slider closer to 6 will result in a slightly sharper, but grainier image. Adjusting
in the other direction (towards 9) will reduce the grain at the expense of
sharpness. Use the Apply button to apply this filter.
Edge Detection
Two specialized edge detection filters are provided, the Prewitt Compass and
Prewitt Gradient. Both of these filters provide similar functionality to their
respective generic filters (described above), however they apply these filters
in all directions. The affect can be quite dramatic, but is sometimes useful.
Clicking the Prewitt Compass button will apply a Prewitt Compass filter for each of eight directions (N,
NE, E, SE, etc.) In effect, all horizontal, vertical, and diagonal edges are
detected. Clicking the Prewitt Gradient button will apply both a horizontal and vertical Prewitt Gradient filter,
thereby detecting edges in both the X and Y directions. This is probably the most
useful edge detection filter.
Degraining
Clicking the Pseudo-Median button applies a very useful degraining and noise reduction filter. This is
a much more sophisticated filter than a simple smoothing function, as it can
completely remove noise. There are three selections for the type of noise to
remove – select either Light, Dark, or All (light and dark). For example, if you have an dark image with a few spots of
random, bright, high-count pixels, applying a light pseudo-median filter can
completely remove these pixels (contrast this to an averaging filter, which
would tend to simply grey and blur them).