Image Processing Services

Geopic

The Image Processing group at MDA has been at the forefront of developing new algorithms since its inception in 1969. Listed below are explanations and examples for some of our most common image processing tasks. See our newest algorithms for specific research and development applications.

GEOPIC is a digital print product derived from any satellite data source. Our image processing scientists filter the data, contrast enhance it and annotate it with a map collar (scale bar, image ID, projection information, north arrow, and title). Clients can specify any scale, band combination, print size or layout. A "Fast GEOPIC" is a low cost, fast turn-around product with a standard contrast stretch and layout. A "Standard GEOPIC" is made according to the client's custom specifications. It can include the client's logo, special grids or graticules, vector overlays, special annotation, multiple languages, etc.

Data Fusion

Data fusion involves taking the best attributes from multiple sensors and merging them into one product. The most common form is fusing a high spatial resolution panchromatic image with a set of lower spatial resolution spectral images.

Fusion Example

For example, using Landsat 7 ETM+ data, the 15m panchromatic band may be used to sharpen the six 30m multispectral bands. This can also be done using different sensors, such as SPOT 10m panchromatic with Landsat 30m multispectral. MDA uses proprietary fusion algorithms that preserve the spectral fidelity of the multispectral bands, may be run on as many bands as needed simultaneously, and allows for some misregistration between the high and low resolution images.

Special Transforms

Spectral transforms manipulate the reflectance information in different spectral bands in order to enhance or suppress features of interest. Below are some of our most popular spectral transforms:

Simulated True Color

To produce a more vibrant image that simulates natural color than, for example, a straight TM band 3, 2, 1 color composite, MDA uses a spectral transform that simulates true color using all six non-thermal TM bands. While preserving the natural appearance of ground features, more detail can be seen within dark, light, and spectrally uniform targets.

Example of simulated true color image

Minimum Noise Fraction

The minimum noise fraction is a spectral transform that is a step beyond principal components. This algorithm is included with the ENVI COTS package. It separates noise from the signal and condenses much of the spectral information into a few bands that may then be used to create a color composite.

TM 742 and MNF comparison

Vegetation Suppression

MDA has developed an algorithm that effectively removes the spectral characteristics of vegetation and reveals the lithology below. More information about this process is available in the R&D section

Vegetation suppression: CA

Vegetation suppression: Brazil

Decorrelation

A decorrelation stretch is a method used to increase the range and diversity of colors within any color image. It is a simple and effective way of improving the visual distinction of an otherwise bland image.

Decorrelation example

Geocoding

Satellite imagery is nothing more than a grid of numbers. In order to know where the image lays on the surface of the Earth, we must geocode the image. This may be done a number of ways, each with its advantages and shortcomings.

Control: DRGs, Maps, GPS or Ephemeris

The map coordinates may be derived for an image by:

USGS
digital raster graphic

• Spacecraft ephemeris (least accurate)
• Map control, either digital (DRG) or paper (accuracy depends on quality of maps)
• GPS points collected at identifiable locations in the image (most accurate)

Image to Image Registration

An image may also be geocoded by finding common tie points (such as road intersections) with an image that is already geocoded. The accuracy in this case is dependent upon the accuracy of the geocoded image and the quality of the tie points. GeoCover Ortho data, which is accurate to +/- 50 meters, is frequently used as a consistent, globally available map base.

Projections and Datums Supported

Geocoded images may then be placed into any projection and datum desired. In addition to an extensive list of standard projections and datums, custom ones may be created provided the necessary parameters.

Orthorectifaction

Looking out the window of an airplane over Washington, D.C., you would notice that the Washington Monument "leans" one way or another, and only as you fly directly over it does it appear as a single point. The "leaning" of the monument is caused by the parallax created by the elevation difference between the base and the top. This parallax also affects topographic features, and causes problems when you try to overlay the image on a map. This distortion may be corrected in a process called orthorectification as long as the elevation of each pixel is known.

Standard DEMs

Fortunately, digital elevation models (DEMs) cover the entire globe and may be used to orthorectify any satellite or airborne imagery. High quality DEMs are available for the United States and much of North America and Europe with resolution as high as 30 meters. For the rest of the world, NIMA's GTOPO30 global DEM is used, which has a resolution of 30 arc seconds (~900 meters at the equator).

Custom DEMs

For areas where the GTOPO30 is not good enough, MDA can derive accurate, high resolution DEMs using SPOT stereopairs. These derived DEMs may be further improved with the availability of accurate base maps to confine the elevation values.

Seamless Mosaics

The keys to seamless mosaics are accurate scene-to-scene rectification, quality orthorectification, and radiometric balancing. MDA has devoted great effort to developing software that does all this.

Massive Block (Bundle) Adjustment

Africa Mosaic

The result of this development is Mospoly^TM. Mospoly^TM is able to block adjust hundreds of scenes at a time, ensuring that the overlap region of adjacent scenes are perfectly matched and orthorectified consistently. This technology can be used to create mosaics ranging from two scenes to mosaics that cover a continent and beyond, and is being used to complete the GeoCover Ortho project.

Hyperspectral

Much of hyperspectral image processing has moved beyond R&D and is now in production status. However, new processing techniques are being continually developed. See the R&D section for more information. The following processing can be rapidly applied to large volumes of hyperspectral data:

• Atmospheric correction
• Post-calibration spectrum enhancement
• Autonomous endmember selection
• Pixel unmixing
• Endmember mapping
• Vegetation stress mapping
• Geocoding
• Aircraft motion compensation

For pricing information, please contact our customer service department.

Digital Products