Featured Image - 01/22/2008
8-bit vs. 14-bit: What a difference 14-bit makes!
The main goal of the Apollo Flight Film Scanning project is to digitally preserve as much of the information content of the original Apollo flight films as possible. To this end scans are acquired with a Leica DSW700 photogrammetric scanner providing very high spatial resolutions (200 pixels/mm or over 5000 pixels per inch) with an extended bit depth (14-bit A/D). Digital images are often digitized as 8-bit pixels allowing only 255 shades of gray. However, the 14-bit capability provides over 16,000 shades of gray thus capturing subtle shades inherent in the original film. The high resolution (over 26,000 by 26,000 pixels) 14-bit images are large (1.2 GB) and many display programs have difficulty successfully displaying these large 14-bit images. Consequently, during the processing for web distribution smaller 8-bit versions are produced for download. From a practical standpoint, although the two versions of the files are very similar to the naked eye, the increased grayscale response of the 14-bit images results in a significant increase in image quality, which is especially useful for scientific analysis.
As an example, below is a small strip of Apollo 15 metric frame AS15-M-0888, showing Tsiolkovsky Crater. The fact that the 8-bit images are limited to 255 gray levels, and that the scene has large contrasts between bright and dark areas, results in some interesting implications for the processing. Figure 1, below, shows a subarea of this image that was enhanced to emphasize subtle changes of brightness in the mare which results in loss of detail in the bright highlands materials in the right.
Figure 1. A section of the Apollo Metric frame AS15-M-0888 showing stretch to enhance mare. [NASA/JSC/Arizona State University]
Figure 2, below, shows a different stretch of this image that more effectively captures the greyscale variation in the highlands. Note detail in the dark mare has been sacrificed. Figure 3 shows a direct comparison between these two images.
Figure 2. A section of the Apollo Metric frame AS15-M-0888 showing stretch to enhance highlands. [NASA/JSC/Arizona State University]
Figure 3. A section of the Apollo Metric frame AS15-M-0888, comparing Figure 1 and 2 stretches. [NASA/JSC/Arizona State University]
For the web 8-bit downloads we have elected to provide a compromise stretch between these two extremes. An example is shown below, in Figure 4, where the subtle variations in highlands and mare brightness are still distinguishable. As a consequence of the applied stretch, some of the grey levels in the extremely bright and very dark regions in this image have been lost-either saturated (in the highlands on the right side of the image) or not distinguishable (in the dark mare regions near the center).
Figure 4. A section of the Apollo Metric frame AS15-M-0888, showing compromise stretch for both highland and mare. [NASA/JSC/Arizona State University]
Although the 8-bit images are outstanding approximations of the 14-bit data, the extended greyscale response in the 14-bit images allows for even higher image quality. Figure 5, below, shows the same image in the full 14-bit version. The 16,000 levels of greyscale response in the 14-bit images (like Figure 5) captures far more of the variation in the bright and dark regions of the image.
Figure 5. A section of a 16-bit version of the Apollo Metric frame AS15-M-0888. [NASA/JSC/Arizona State University]
Figure 6, below, shows the advantages of the 14-bit images in a different way. The digital numbers (DNs) from a line profile through the center of the 14-bit image are plotted on the x-axis; the DNs from same profile across the 8-bit image are plotted on the y-axis. This plot illustrates two points. First, the 14-bit image has almost 16,000 greyscale levels covering the middle range (the solid line, below) which means a substantial increase in image quality. Second, the ellipse highlights the saturated pixels in the highlands on the right side of the 8-bit images. Although the 8-bit image is capped at 255 grey levels, the 14-bit image can capture the bright pixels that are being lost in the 8-bit image, providing for finer detail and higher image quality.
Figure 6. XY Plot showing distribution of DN values comparing 8bit and 14bit images. [Arizona State University]