Radiometric Enhancements
Introduction
In the previous section we covered radiometric errors and atmospheric correction techniques. Radiometric enhancements differ from the pre-processing techniques in the fact that they are primarily designed to improve or alter the appearance of the image. These methods can enhance subtle radiometric differences so that the eye can easily perceive them. Radiometric enhancements manipulate the digital numbers in the images and change how the image or data is displayed on the computer screen. Many of these techniques use the image statistics and histograms to produce the intended results.
- Contrast Stretching
- Level/Density Slicing
- Image Thresholding & Masking
Contrast Stretching
As we learned earlier, computer monitors have a range of brightness from to 0 to 255 (8-bit) and often pixel values in raw images don't fill the full range or may have a range greater than 256. Contrast stretching improves the contrast of the image by “stretching” the pixel values ( or DNs) to take advantage of the full range. Most image processing software automatically apply a contrast stretch when displaying images. It is important to note that most contrast stretching does not change the actual pixel values of an image, it simply changes how it is displayed.
Linear Contrast Stretch
Linear contrast stretch linearly expand the original pixel values of data into a new distribution. There are several methods of linear contrast enhancement, the most common are Minimum-Maximum and Percent Linear contrast stretch. In the minimum-maximum linear contrast stretch, the original minimum and maximum values of the data are assigned to 0 and 255 respectively and all other values are stretched evenly between. The percentage linear contrast stretch is similar to the minimum-maximum linear contrast stretch except this method uses a specified minimum and maximum values that lie in a certain percentage of pixels from the mean of the histogram.
Histogram Equalized
Histogram Equalized contrast stretch assign more display values depending on the frequently of these values. This way there is more detail in areas with the greatest frequency of pixel values. These areas will be better enhanced relative to those areas of the original histogram where values occur less frequently.
Other Contrast Stretches
There are a variety of other contrast stretch types. The Standard Deviation Stretch exclude the extreme values in the histogram, then apply the linear stretch to remaining values. There are also Selective stretching techniques that allow the user to select a custom select range of values to stretch. another common contrast stretch is the Gaussian stretch, which fits the observed histogram to a normal, or Gaussian, distribution.
Density/Level Slicing
Density or Level Slicing divides up raster pixel values into a series of intervals or “slices”, with different colors applied to each slice. This technique is often performed on single band images to highlight differences in values. The process essentially takes continuous data values and divides them into discrete classes or slices based on histograms or user specified values. Most density or level slicing operations use histograms and image statistics to create the levels or slices.
Most geospatial analysis programs allow the users to specify the number of slices/classes and control the range of values in each class. In some cases, users might want manually specify the number of classes and range of values based on the specific data. Histograms and image statistics can also be used to automatically divide or slice the data into classes. For example, equal interval methods divide the range values (based on the minimum and maximum value) into equal-sized ranges. Natural Breaks classes are based on natural groupings inherent in the data. Class breaks are identified that best group similar values and that maximize the differences between classes. In the quantile method each class contains the same number of pixels.
