Technotes
New and Improved Digital Orthophoto Imaging Techniques
By David Brostuen

The combination of modern computing horsepower and advances in digital imaging technology are resulting in exciting developments in producing digital orthophoto imagery (DOI). Practical solutions have been developed for many of the problems once considered "inherent" to DOI products. Historical mosaicking problems such as distortion of buildings along image boundaries and the checkerboard effect between images due to color inconsistencies are rapidly diminishing. Sophisticated processes for color correction are reducing human subjectivity and bias resulting in imagery that closely resembles reality. Effective methods for removing traditional problems such as leaning buildings and distorted bridges are enabling production of DOI with unprecedented accuracy. Combined with these advancements, the availability of robust computers for large volume data processing is collapsing the run-time of previous DOI processes from hours to minutes. Experienced DOI users have realized these technology gains and have dramatically increased their expectations regarding affordability, accuracy, and overall image quality.
    Digital orthophoto imagery is commonly used as a source for capturing various thematic data layers and as a visual backdrop in a GIS. Cities, counties, and utilities have found DOI to be a powerful visual quality control tool in assessing the positional accuracy of features including streets, parcels, manholes, and utility poles. Natural resource, engineering, and planning organizations use digital orthophotos for accurate resource assessment, change detection, and inventory studies. Non-traditional users such as real-estate companies are gaining added value through integrating DOI with visual information on school districts and proximity to hospitals, shopping, and other services. The use of DOI is beginning to become more widespread by the general public due to increased resale activities and low-cost imagery viewers.
    Modern DOI is aerial or satellite imagery that has been processed to remove effects including terrain variation and camera distortion, then rectified to a map coordinate system. Aerial imagery may be derived from either scanned aerial photography or a digital camera/sensor. The orthorectification process requires development of a digital terrain model (DTM), and accurate ground control and calibration information to eliminate the effects of terrain variation and camera/sensor geometry. Once rectified, the image tiles are commonly mosaicked, then balanced to produce consistent color and tone across the project area.
    Although advances in quality and accuracy have been significant, the methods required to correct many of the traditional problems can be very time consuming. The primary reason is that, despite using modern computer processors, many of the procedures have required significant human interaction and input. For example, in order to produce a "seamless" image mosaic in an urban area, a seam line must be defined between adjacent images that avoids buildings, bridges, and other features. In dense urban areas, delineating seam lines manually may comprise the single most time consuming task in the total DOI process. Only now are semi and fully automated techniques being developed to address the most laborious tasks required to produce the current generation of DOI products.
   Analytical Surveys, Inc. (ASI), has re-engineered its DOI production process to both tackle traditional DOI problems and take full advantage of modern computer processing power and technology. The new process, called the Method for Elimination of Tilt and Relief Displacement in Orthophotography (METRO), is a fully automated technique developed by ASI's Advanced Technology Division, based in Colorado Springs, Colo.. METRO was originally designed, in part, to provide an effective method to address the problem of relief displacement of features above the ground such as buildings and bridges.
    According to Anthony Thorpe, ASI's vice president of Research and Development, "METRO is the culmination of 2 years of research and development at ASI." Initial development focused on producing True Orthoimagery using fully automated procedures. In a "True Ortho," typical problems due to relief displacement are corrected. The most obvious benefit is the removal of building lean and correcting warped bridges. Recent advancements in METRO have focused on improving both the quality and efficiency of the mosaicking process. Breakthroughs in automatic seam line delineation and color correction have enhanced the capabilities of METRO and automated the most laborious and subjective parts of the production process.

Mosaicking Pitfalls
Mosaicking digital orthophoto imagery typically involves merging individual images into a single large file, then performing a feathering routine to blend adjacent images along the edge in an effort to make it appear "seamless." The process is relatively effective in areas of sparse development and low terrain relief. Feathering processes are commonly used in commercial DOI software to "blend" seams. However, when consideration is not given to where seam lines are placed between adjacent aerial photos, mosaicking problems arise. One example of image artifacting is known as "ghosting." In this case, the problem is caused by the feathering of imagery along a seam line. Excessive ghosting can cause significant loss of information and distortion, particularly in urban areas where extensive detail is required.
    Another problem occurs along seam lines between two aerial photos with buildings and other features that have significant relief displacement. Figure 2a shows what can occur if a seam line (shown in red) bisects a building. Notice how the building is distorted. A portion of the building is missing due to the difference in relief displacement between adjacent images. Both relief displacement and ghosting can be avoided if a seam line is routed around buildings and other features, rather than through them. Some of the currently available DOI software products have a capability to manually delineate a seam line between two adjacent photos. However, manual seaming is an intensive process that can consume a significant amount of time and effort, and requires careful delineation of each seam by a skilled technician.

Auto-seaming
ASI's Advanced Technology Division (ATD) has developed a process that automatically routes seam lines around buildings and other features in a batch-processing environment. Figure 2b shows a seam line that was generated through the automated process using METRO. Note how the red seam line has avoided the building by routing the line down a nearby road. The result is a visually seamless image that is free of noticeable seam lines and displacement problems.
    Since the seaming process is highly automated, little human intervention is required in producing the seam lines. The process generates non-linear seam lines that significantly reduce or totally eliminate the need to run feathering algorithms to blend seams between adjacent imagery. As a result, artifacts such as ghosting are also greatly reduced or eliminated. Delineation of the seam line is based on differences in radiometry between adjacent pixels.
    After the automatic seaming program is completed, imaging technicians perform visual checks to verify the acceptability of each seam line prior to mosaicking the imagery. Seam lines requiring further adjustment can quickly be modified at this time. According to Thorpe, "In producing top-quality digital ortho-imagery, a significant cost is attributed to defining seam lines manually. With METRO automated seaming, ASI again leverages computer technology to build better products for lower cost." The result is mosaicked imagery that appears more seamless and requires less technician time and effort compared to manual seaming techniques.
    ASI has structured their DOI production to execute the auto-seaming programs during off-hours, thereby capitalizing on the maximum CPU time and reducing production costs. The auto-seaming program run-time varies based on the resolution of the imagery and file size of the digital imagery being processed.

Hands-off Color Balancing
Differences in color and contrast between adjacent aerial images are a common occurrence. More often than not, the problem is due to an inability to fly an entire mission and acquire all of the imagery for a large project area in a single day. Many of ASI's projects involve several thousand individual photographs, at times, acquired over several weeks. Variations in atmospheric conditions and sun angle result in noticeable differences between adjacent flight lines of imagery. Other factors attributing to color disparity include variation in film processing and scanning methods (for aerial film) and reflectance characteristics such as hot spots. A hot spot is due to the reflectance of sunlight into the camera lens. The result is a washed out (overexposed) bright spot where much of the terrain detail is lost.
    To reduce subjectivity and human interpretation, ASI has developed sophisticated automatic techniques for balancing the color of mosaicked digital imagery. ASI's Interactive Color Editor (ICE) software begins with calculations that are made during the rectification process. In order for the software to adjust the images, the position of the sun in relation to the camera has to be determined. This enables the software to automatically adjust for hot spots and reflectance effects caused by ground cover, terrain, and sun position. Figure 3a shows several images for an area in El Paso County, Colo. Notice the variation in color between each image. Figure 3b shows the same images after automatic color balancing software has been applied using METRO.
    At times, color balancing digital orthophoto imagery is a highly subjective process based on human perception of what a person believes looks "good." People sense color differently, what appears acceptable to one person, may not be pleasing to another. The phrase "beauty is in the eye of the beholder" has never been more true than when applied to digital imagery. ASI's ICE software greatly reduces the subjectivity factor in the color balancing procedure.
    ASI's color correction process is a two step process. First, the software performs image sampling on several patches of ortho-imagery to identify color correction parameters. As a quality control measure, ICE allows the user to either manually adjust settings (interactively) or create a "target" image and let the software automatically determine and adjust settings. The imaging technician then views the imagery with the correction parameters to ensure effective color balancing will result. However, according to Wolfgang Schickler, ASI's ATD manager, "Because of the tremendous success of the automatic capability, manual adjustment is now hardly ever used." The actual color correction occurs in the mosaicking process.
    The software can be applied at several stages throughout ASI's DOI production process to balance imagery within delivery areas, as well as over an entire project area. Like the auto-seaming process, significant productivity gains are realized by running the color correction programs in a batch-processing environment.

Conclusion
Modern advances in computer technology and photogrammetry are both changing the way digital ortho-imagery is produced and improving the integrity of the product. New automated mosaicking processes are improving production efficiency and affordability while removing subjectivity by applying more scientific principals, rather than individual human bias. Near-term gains in DOI production will be driven by technology advancement in digital airborne camera systems and integration of multiple sensors such as LIDAR and interferometric radar systems. High resolution satellites offering world-wide coverage and revisit capability will soon be available for producing 1m DOI. Common problems resulting from aerial cameras such as pressure scratches and lint will not exist in the satellite imagery, although clouds may be present. Color aerial imagery and multi/hyperspectral digital imagery demands will increase, driven by the low cost of data storage and modern compression tools. Internet access, affordable "pre-packaged DOIs," and user-friendly viewing software will inevitably expose the product to new users more quickly than ever before.

About the Author
David Brostuen is the director of Digital Imaging at Analytical Surveys, Inc. More Info Analytical Surveys, Inc., 1935 Jamboree Drive, Colorado Springs, CO 80920.

Back