Landsat Haze Reduction

As solar light traverses the atmosphere it is scattered. One type of scattering is called Rayleigh scattering. In this type of scattering, electromagnetic radiation interacts with matter whose size is much smaller than the wavelength of the EMR. Haze can also be caused by atmospheric aerosols. An aerosol is a suspension of fine particles or liquid droplets in a gas. Examples are smoke, oceanic haze, air pollution, clouds, soot and photochemical smog.

Short wavelengths of visible light (mainly violet and blue) interact are most affected by Rayleigh scattering, which is inversely proportional to the fourth power of the wavelength, Longer wavelengths (red, near infrared (NIR), infrared (IR) etc. are relatively unaffected and pass through the atmosphere without interacting at a molecular level. A second type of scattering called Mie scattering affects longer wavelengths of light. In this type of scattering, electromagnetic radiation interacts with much larger aerosol particles and can be scattered in the same way that shorter wavelengths are scattered by gas molecules. The wavelengths affected depend on the size, shape and texture of the particles encountered by the radiation.

The top of atmospheric (TOA) radiance sensed by orbiting satellites is the result of a complicated series of interactions between the EMR, the atmosphere and the ground both downward toward the earth from the sun and from the earth to the satellite sensor. This is shown in the Figure 1 where L1 is the atmospheric radiation, L2 is the reflected radiation and L3 is adjacency radiation. L4 and L5 represent backscattered radiation that represents a net loss of available sensed radiation. Only radiation component 2 contains information from the currently viewed pixel. Normally, the task of atmospheric correction is the removal of components 1 and 3 and the retrieval of component 2.

Since only a portion of EMR reaching the satellite sensor is actually reflected from objects on the surface under observation, the result is a blurring effect, often significantly limiting the view of surface features. The result of both types of scattering is less of the incident light reaching the surface and more radiation received at the satellite sensor from the atmosphere.

There are a few widely-used techniques used to reduce the effect of molecular and aerosol scattering. One technique is called the Dark Image Adjustment. This is a histogram-based method that determines the onset of the lower edge of the histograms for a band histogram. It then subtracts this value from each digital number (DN) of the image. If the subtraction causes the DN to be less than zero, it sets it equal to zero. In addition to reducing the haze component of a band file, it of course decreases most of the DNs of the image, thus the name.

PANCROMA has a preprocessing utility for haze reduction of an RGB image using the Dark Image method. To use it, first open the three component band files by selecting 'File' | 'Open' from the menu. When the three files are open, select 'Pre Process' | 'Haze Reduction' | ''Dark Image Technique'. After the files are opened, you will be prompted to input a cutoff value. The cutoff is the minimum number of DNs in a channel that determines the location of the threshold. For example, if the cutoff is 500, and channel 44 of the histogram is the first one to have at least 500 DNs in it, then the threshold and therefore the onset of the lower edge of the histogram will be determined to be 44. The figure below illustrates the idea.

You must specify the cutoff value using the 'Dark Area Data Input' form that will appear after you specify your band files. The form is shown below.

The default cutoff is 500. You can adjust it from zero to 2000 using the slider bar. If you check the 'Compute Cutoff' check box, PANCROMA will compute the cutoff value automatically. The images below show the results of applying the Dark Image haze reduction algorithm. The first image is the unprocessed image. The second has had the haze reduction applied.

There is considerable reduction in the haze and also the blue tone due to the Rayliegh scattering effect.