Image Enhancements Overview
Introduction
Image enhancements are performed to improve the visual interpretability of images and to improve analysis. There are three main types of image enhancements, radiometric, spatial and spectral. In this section we will focus on the first two, radiometric and spatial.
Histograms and Image Statistics
Histograms
Before we begin reviewing enhancement techniques, we first must understand the concept of an image histogram. A histogram is a graphical representation the number of pixels in a single band corresponding to each possible pixel value. The brightness values (i.e. 0-255 for an 8-bit image) are displayed along the x-axis of the graph. The frequency of occurrence of each of these values in the image is shown on the y-axis.
Histograms used for image enhancements including: contrast stretching, atmospheric correction, and other statistical manipulations. In multispectral imagery, there is a hisogram for each of the bands.
The image on the left is a Landsat 8 image of Lake Tahoe. The chart on the right is a histogram that shows the distribution of pixels in this image. Each of the colored lines represent the distribution of pixels for that particular band. For example, the red line represents the distribution of pixels in the red band. Note the spike of blue pixels towards the left side of the histogram. This spike represents the large number of water pixels in the image that have similar values in the blue wavelengths.
Statistics
In order to perform many of the image enhancement techniques, statistics for rasters must be calculated. These statistics can include the following:
- Pixel count and percentage per class for classification images.
- Basic statistics (minimum, maximum, mean, and standard deviation of values) for single- or multi-band images.
- Histograms with frequency distributions (number of points, cumulative points, percent for each bin, and cumulative percentage for each DN (digital number)