By design, JPEG 2000 offers no specific support for any particular application domain, such as georeferencing metadata for geospatial imaging. This means JPEG 2000 doesn’t specify mechanisms for georeferencing the image, describing the sensor model used to collect the data, or correlating features within the imagery to other GIS datasets. For good reasons, the standards committee decided not to support such domain-specific requirements. Instead, the committee left room within the JP2 file format for “boxes” containing arbitrary XML data that can refer to the image data within the file.

Enter GML
The Geography Markup Language (GML) is an XML grammar used to describe geographic data such as coordinate reference systems and positioning, geographic features, sensor models, annotations and styling, etc. Like JPEG 2000, GML version 3.1 will be an ISO standard (ISO 19136) in its own right.
 

Despite its richness, GML doesn’t provide for “metadata” in the normal sense of the term, such as Federal Geographic Data Committee (FGDC) metadata. GML is a language used to construct definitions for features, geometries, etc.; it doesn’t define any concrete features on its own. GML is used to construct XML application schemas for use within a given application or system. GML does allow for inclusion of other XML data formats and referencing of other entities, possibly external to the file itself; for example, FGDC metadata is explicitly allowed.

Consider the simple case of an aerial image with an associated coordinate system and position information. One can easily envision the coordinate system and position being represented in GML and stored within the JP2 file using the allowed XML boxes (see Figure 1). On the other hand, at least the positioning data could be represented with a traditional 6-line world file (.wld, or perhaps .j2w). This wouldn’t define the coordinate system, although other header file formats could certainly be used or invented for use with JP2 files.

Why Use GML?
It seems simple enough to express a world file expression. So why should users go all the way to using GML? Consider three more complex situations, which show how the power of JPEG 2000 can be exploited using GML.

Sensor Models
Original satellite and aerial imagery are now much more than just three 8-bit RGB bands, and JPEG 2000 supports many of the imagery requirements users now see, such as bit-depths of 16 or higher and hyperspectral bands. These images typically have rich sets of associated metadata. For example:

• Full descriptions of the cameras may include sensor characteristics such as the number and wavelengths of the spectral bands, precision and calibration information, and type of sensor (pushbroom, etc.)

• Positioning information may include the usual coordinate system and position of the image relative to Earth, but also may include camera positioning information such as camera angle, date and time, orbit track, etc.

• Image quality information might include cloud cover estimates, NIIRS rating and air quality at time of collection.

Using GML, an application schema can be written that captures this metadata structure; this could be defined and provided by the imagery vendor (see Figure 2). Within their supplied JP2 imagery, GML instance data would provide the actual metadata content. The application schema doesn’t even have to be resident in the file, but could be hosted at the vendor’s or satellite company’s Web site via the standard XML linking and referencing mechanisms.

Multiple Images and Feature Identification
JPEG 2000 permits multiple images to be contained within the same file. Consider a workflow in which images of a particular region are to be captured and analyzed during a period of months:

• Multiple sets of stereopair imagery are to be stored.•
Each pair has associated metadata, including date and time stamps.
• The images all share a common coordinate system, but all are at slightly different coordinate positions.
• Any individual image may have specific features identified on that image.

All of this information may be contained in a single JP2 file. In addition to the data corresponding to the individual images, the file may contain “shared” GML data describing the coordinate system and feature descriptions for all images and “private” GML data for each image describing specific feature instances and offsets within the image (see Figure 3).

The Spatial Web
Increasingly geospatial data are being used in the “Spatial Web” environment, where Web services are used to transparently perform operations such as:
• Providing catalog access to large archives of geospatial data.
• Exporting the data itself, based on query parameters.
• Performing mosaicking and layering operations.
• Performing simple feature classification, description and extraction.
• Styling the data for presentation.

Such workflows are already in use today with vector data; GML is the “language” of the GeoWeb used to describe regions and extents, define and label features and express queries. As users add raster imagery to this system,

Consider Department of the Interior (DoI) imagery and two agencies that rely on the imagery: the Forest Service and the National Parks Service. The Forest Service may wish to view an image with forested regions accentuated through false color imaging, while the Parks Service may wish to view only the visual spectra of the same image. Using JPEG 2000, DoI creates and stores a single image—annotated with GML for the classifications—that can be styled by a Web service according to particular viewer needs (see Figure 4).

Bridging the Gap
JPEG 2000 is coming on fast—vendors are investigating its use, and some already support the standard—but so far JP2 is viewed simply as a “new compression format.” JPEG 2000 is much more than that; it offers the geospatial community the opportunity for a single, common, well-standardized format for most imagery needs during the coming decade. And with its support for networked environments, it’s perfectly aligned with the vision that groups such as OGC have for future networked GISs.

Still, getting there will require some work. The JPEG 2000 standard doesn’t say how to mosaic an image with other data projected to some common ellipsoid, nor does it help users label a region of the image as “a two-lane, paved road.” Fortunately, the standard does allow users to leverage another new ISO standard, GML, to bridge the gap between raw imagery and existing geospatial systems.

Publisher’s note: To read more about GML and JPEG 2000, visit the
following Web sites:
• www.jpeg.org—the Joint Photographic Experts Group
• xml.coverpages.org/geographyML.html—a good GML Web site hosted by the Organization for the Advancement of Structured Information Standards (OASIS), a non-profit, international consortium that drives the development, convergence and adoption of e-business standards.
• www.opengeospatial.org—Open Geospatial Consortium Inc.

 

  See  Featured Images
 
  Subscribe to Earth Imaging Journal

  Go to Home Page