Current Issues
Archives
Media Kit
Editorial Guidelines
About Us
Contact Us
Subscribe

 

 


HOME > ARCHIVES > 1995 > JUNE
Observations: Scanning and Digital Imagery
By G. Michael Ritchie and Mark E. Meade

While digital imagery has become the rage of the '90s, its benefits can potentially be negated by a lack of understanding regarding the scanning process - in particular the appropriate resolution at which to scan original images.
      Although the use of digital cameras for image capture is growing, by far the most common equipment used by photogrammetric firms today is high-resolution cameras which capture images on negatives. A scanning process is then used to convert the photographic image on the negative (or a diapositive copy made from the negative) into digital raster data.
      By definition, a raster image is an image comprised of small, independently controlled pixels arranged in columns and rows. Each pixel is a representation of a specific ground area and is assigned a single value (color or gray scale) based on the predominant color of that area as determined during the scan. The resolution of the raster image is determined by the user at the time of the scan by specifying the density of these pixels.
      While the hardware technology of scanners has improved to allow very precise scans, considerable thought must be undertaken before arriving at the proper scan resolution for a project. Although more precise resolutions are desirable from a visual aspect, file size and the availability of output options must also be considered. For example, the large files which result from ultra-precise scans perform poorly on most hardware platforms. In addition, although precise plotters are available, it would be overkill to perform an ultra-precise scan if the output is to be performed on a plotter with far less resolution than the image itself.
      Thanks to the popularity and affordability of desktop scanners, most everyone is familiar with the specification of dots per inch (dpi). However, precision photogrammetric scans are generally specified by the metric unit of microns. It is simple to convert from one unit of measurement to the other as one micron is equal to one thousandth of a millimeter.
      Whether talking about microns of dpi, all the same principles apply. The resolution is simply a linear measurement taken along one side of a rectangular pixel of information. In the case of dpi, simply divide 1 by the scanning specification to get the measure (in inches) of this dimension. For example, an 800 dpi scan would be composed of square pixels 0.00125 inches (1/800) on each side. A 25-micron scan would be composed of pixels measuring 0.025 millimeters (25/1000) on each side. For comparison purposes, a 25-micron scan is equivalent to 1016 dpi.
      The resulting file size can be easily calculated when given the area, scanning resolution, and format. When scanning black and white imagery, the resulting scan is typically expressed by 256 scales of gray, ranging from black to pure white. Each pixel is assigned one level of gray which requires 8 bits of computer memory - or 1 byte. Color scans require more computer memory to accurately represent pixel colors. Most color scans are represented by pixels comprising 24 bits of information, or 3 bytes; this provides a palette of 16.8 million colors.
      While the hardware technology exists to produce precision scans to 5 microns, the resulting file size from such a scan would be enormous and result in unsatisfactory performance on all but the most powerful computers available today. For example, a 9" x 9" contact print scanned at 5 microns in black and white would result in a file size of approximately 2 gigabytes! It is important to note that because scanning resolutions are expressed as linear measurements, file sizes which are dependent on area vary with the square of the ratio of the scanning resolution. In other words, the same image described above scanned at 25 microns would be 80 megabytes -2 GB x (5/25)2.
       The original scan determines the maximum resolution that can be attained from the digital imagery later - regardless of the enlargement of the final mapping performed from the original photography. In digital imagery each pixel represents a rectangular area on the ground, therefore, this resolution places an upper bound on any imagery developed from the original scans.
      If maintaining an exact image area is necessary, it is important to select a resolution whereby the total image area of a single sheet is evenly divisible by the ground pixel size. For example, a 1"=500' plot with an area of 20 by 30 inches (10,000' by 15,000' image area) would work well with 1-, 2-, 4-, or 5-foot ground pixels. However, a 3-foot pixel would result in a fractional pixel in one direction (10,000/3 =3333.3) and would therefore be inappropriate under this scenario.
      Table 1 lists the ground-equivalent pixel size at different photo scales and scanning resolutions. It should be noted that original scans can be resampled to produce a resolution less than the original image, however, the only way to gain more resolution is to rescan the original image. This table illustrates why it is possible to develop 4- or 5- foot ground pixels from 1"=2000' photography scanned at 50 microns, but a ground pixel of 2 feet could only be developed from a new scan at a finer resolution - in this case 25 microns.
      The flying height and resultant photo scale are important selection factors when planning a project. A 12.5-micron scan of a 1"=2000' image will not produce the same results as a 25-micron scan of a 1"=1000' image, although it would seem so mathematically. In fact, the 25-micron scan will produce better imagery because there is less degradation of the original image at a lower altitude from atmospheric factors, limitations of the photographic film, and the resolving power of the camera.
      A thorough understanding of the scanning process is important to arrive at the appropriate scan resolution. Careful planning will ensure a final product that meets the needs of a particular project while maximizing the speed at which the image can be manipulated on a given hardware platform.

About the Authors:
Mike Ritchie is president, and Mark Meade is vice president of Photo Science Inc., of Lexington, Ky. They can be reached at 606-277-8700.

Back

 

©Copyright 2005-2021 by GITC America, Inc. Articles cannot be reproduced,
in whole or in part, without prior authorization from GITC America, Inc.

PRIVACY POLICY