By Michael P. Gerlek, software architect, LizardTech (www.lizardtech.com), and Cody Benkelman, consultant, Benkelman Engineering (benkelman@yahoo.com).

Humans couldn’t exist on Earth without the sun, and we couldn’t fly above Earth and capture images of it without its unique atmosphere. Ironically, the sun and atmosphere also wreak havoc with aerial photos and satellite images by creating tonal imbalances, as well as imbalances from one image tile to another in image mosaics, that degrade the visual quality of orthoimagery. Although some tonal imbalances are only cosmetic, others reduce an image’s
usefulness.

Why Tonal Balancing Matters
Correcting tonal imbalances is important for several reasons. A balanced image is more pleasing to the eye, and a user can more readily subordinate it mentally to other layers of information. On the other hand, an image with uncorrected tonal errors is a distraction at best and an unusable product at worst.

Many government agencies that use orthoimagery “downstream” will not accept tonal imbalances beyond a certain point. For example, the U.S. Department of Agriculture (USDA) National Agriculture Imagery Program (NAIP) requires tonal balancing on all delivered imagery.

Imagery is sometimes corrected before distribution by data providers, but not always, and even in these cases imbalances can persist because the corrected imagery may not match data that an agency already has from other sources or other years. This highlights a key challenge in correcting tonal imbalances: In most cases, there is no objective “right answer,” so tonal balancing is a subjective process by nature.

The examples below illustrate three specific tonal balancing issues, but such concerns take many forms and appear for many reasons. Some only become problematic when images are “tiled” together to create a mosaic. The catalog of color-balancing horrors includes numerous entries (see Common Causes of Tonal Imbalances,” below).
 

Grouping the Issues
The visual imbalances that bedevil geospatial imagery users can be divided among three broad categories: spatially uniform, spatially nonuniform with local imbalances and spatially nonuniform with global-trending imbalances.

Spatially uniform imbalances: Any imbalance that affects an entire image or image tile without variation is said to be a uniform imbalance. Examples would be images that are too dark or too light, have too much or too little contrast, or have an undesirable colorcast. Note that imbalances may not even be noticeable (nor even considered an error) if the image is viewed separately. Although each tile composing the mosaic (A) above is uniform, the mosaic is imbalanced.

Spatially nonuniform local imbalances: These imbalances affect only a portion of an image or are otherwise varying in their effects on an image, as illustrated in image B above. Cloud shadows and sun glinting off water can create locally nonuniform imbalances. Such imbalances are one of a kind and corrected individually when correction is possible.

Spatially nonuniform global-trending imbalances: The varying ways Earth’s surface reflects sunlight, the direction and angle of the sun in relation to the sensing equipment, and atmospheric phenomena such as scattering can create imbalances that are neither local nor completely uniform across an image, as shown in image C above. Although these imbalances aren’t simple to model, they show a repetitive pattern in sequential images that lends itself to correcting imagery in a group.

If Sam must compile a mosaic product from mismatched data, any imbalances, which are almost certain to be present, become pronounced. A single image tile, or a set of tiles, may look fine until placed adjacent to others that are brighter, darker or have a different colorcast.

Let’s say Sam wants to update an existing mosaic originally composed of data that wasn’t acquired from NAIP—perhaps the images were taken in a different season or year, or they were created from faded film originals. To imagine the simplest example, Sam wants to combine two new NAIP tiles and two older images to create a four-tiled image mosaic. Although each image may look fine on its own, tonal variations between tiles will be pronounced when they are mosaicked together, as shown in the image below. When this occurs in a mosaic with 25-100 tiles, the result is a distracting, patchy composite.

Correction Tools
The range of tools available to correct tonal imbalances is broad, but not particularly dense. Several high-end tools are available (some as integrated toolsets within photogrammetric applications), providing some degree of success at correcting tonal mismatches within a cohesive dataset.

On the comparatively inexpensive end, popular photo-editing tools are used extensively to painstakingly touch-up individual nonuniform local imbalances, and even some batch processing, but such tools don’t handle geospatial metadata or extremely large images, nor do they allow intelligent “neighbor image” analysis and processing typical in a mosaicking environment.

Another option is Version 6.0 of LizardTech’s GeoExpress software. If Sam already uses GeoExpress to convert the raw imagery to industry standards such as MrSID (.sid) or JPEG 2000 (.jp2), only a few additional steps in the user interface are necessary to correct uniform errors. Sam’s workflow will involve pulling image tiles from new, color-balanced county NAIP data and older GeoTIFFs into the GeoExpress user interface, performing the desired color-balancing operation, and outputting a new mosaic.

LizardTech also is helping to create the nascent GMLJP2 standard, which enables much richer, GIS-oriented metadata than was possible before. GML metadata can hold information such as a formal, vendor-supplied description of the camera or sensor model used in image capture, as well as environmental and other data, including position, height and angle of the sensing equipment and the time and date of capture. GML metadata also can store information about any corrections that have been applied to an image set—this is critical information for those using brightness values within imagery for measurement.

No matter what tools may have been applied prior to mosaicking, the end user must judge whether the full product is adequately balanced. Although correcting tonal imbalances is complex, today’s geospatial professionals have an expanding range of tools to correct tonal imbalances.
 

 

  See  Featured Images
 
  Subscribe to Earth Imaging Journal

  Go to Home Page