1000 PPI: Method or Maddness

I wanted to clarify the issue regarding the 1000 PPI requirement because the NIST requirement for a resolution of 1000 PPI is both confusing and misleading.  First, the 1000 PPI requirement is technically an output requirement and has little (or at least nothing good) to do with image capture (I'll explain more about that below).  The reason why the 1000 PPI requirement exists is to provide a standard for interfacing with systems such as AFIS, which is why this is so grossly misunderstood. 

The 1000 PPI resolution is based on a physical size requirement to ensure that all systems can size the image to occupy the same area size.  For example, if you have a one-inch wide image with a resolution of 1250 PPI and submit that image to AFIS, the AFIS would resize the image, making the image 1.25 inches -- the first 1000 pixels representing one inch, and the remaining 250 pixels representing an additional 0.25 inches. As a result, the latent image would not match the corresponding tenprint image if it existed within the AFIS database. So if the latent print resolution is set to 1000 PPI before exporting the image to the AFIS, all images will be sized properly and one inch would always represent one inch. (A one inch image with a resolution of 1000 PPI would appear as two inches on a system at 500 PPI, where the first 500 pixels would represent the first inch, and the second 500 pixels would represent a second inch.)

If you have a scanner that has an optical (hardware) resolution of 1600 PPI and you set the scanner to capture 1000 PPI, you are actually compressing (even though they call it resampling) the image, thus significantly reducing image quality. The reason why this is compressed is that the scanner actually captures 1600 values. Since it has to eliminate a significant number of pixel values, it averages the pixel values together. (With a 1600 dpi scanner, you are actually capturing a total of 2,560,000 pixel values within a one-inch square; using a 1000 dpi setting, you are only saving 1,000,000 pixel values within that same one-inch square. Since the scanner cannot just arbitrarily throw out 1,560,000 pixels, it resamples the image using the nearest neighbor resampling technique, which averages two or more color values together, i.e., black plus white equals gray to reduce the number of pixels (2,560,000 pixel values to 1,000,000 pixel values) it sends to the receiving application, such as Photoshop.)

If the opposite is true let's say you have an image with a resolution of 850 PPI and the image appears clear on the screen and it is clearly identifiable (i.e., you can clearly see both level 2 and level 3 detail), then you can simply adjust the image resolution to 1000 PPI (in Photoshop with the Image Size dialog box, making sure the Resample Image checkbox is checked and the resample option is Bicubic or Bicubic Smoother).

Having a higher image capture resolution really only impacts two separate issues: (1) are you going to zoom into the image to view the level of detail at a greater magnification (and avoid pixilation at a lower magnification)? And/or (2) are you going to be doing any image enhancement (processing) on the image.

As you well know, the higher the image resolution, the more detail you can see when you zoom into the image. When you are viewing an image on your monitor in Photoshop, the system shows the percent (sampling) of the total number of pixels contained within the image. For instance, when you look at an image in Photoshop, you will see the percent of actual pixel values (sampled) at the top of the image window.

To see the full image detail (resolution), you would have to zoom into the image until you see 100% displayed on the image title bar. When viewing 100 percent of the image pixels, you cannot view the entire image on your screen ¦ as you can only see the total number of pixels that can be displayed on your screen. (If your monitor has a setting of 1024 x 768, you can only see 1024 pixels across the full width of the screen. Approximately 150 pixels are used for the tool bar and palettes in Photoshop (going from left to right on your Photoshop window), which leaves a maximum of approximately 865 image pixels. (You would have more screen pixels to display more image pixels if you hide the toolbar and palettes by hitting the tab key, which is a toggle switch to turn on and off the toolbar and palettes.)

The higher resolution REALLY is important when you are trying to enhance an image, such as using color channels to eliminate background noise or the pattern removal filter, or if you are using the Shadows/Highlights function to adjust the contrast between faint ridges and darker furrows. While we (the human) may not be able to make much sense when viewing the image at pixel level, the computer is actually performing its processes at that level. So when you resample images, either through file compression such as JPG or resampling as discussed above with the scanner or simply by moving the camera further away from the object thus requiring one photo diode to capture a larger area thus sampling the color values it sees to obtain a single pixel value, you are actually combining two or more color values to make a single averaged color value. When you average those values together, you are actually eliminating the possibility that the computer can effectively enhance the image ¦ in other words, the merged ridge detail and background cannot be separated and enhanced.

Using another example to illustrate this point, you must have five columns of pixels to capture a single pair of lines with sufficient level of detail. One column of white pixels for the white space on the left, one column of black pixels for the left line, one column of white pixels for the white space separating the lines, one column of black pixels for the right line, and one column of white pixels for the white space on the right side of the right line.

If you have only four columns of pixels (i.e., you move the camera back to cover a larger area and only have four columns of pixels to cover the space occupied by the line pair), you might get the following result: one column of pixels for the white space on the left, one column of black pixels for the left line, one column of gray pixels (one column of pixels now must sample for both the white space separating the lines and the right line, which is sampling both black and white, thus giving you gray), and one column of pixels for the white space on the right side of the right line. As a result, you no longer have two separate lines; you have a black line with what appears to be a gray shadow, thus you have lost image detail.The bottom line is that you should capture images at the highest possible resolution that you can. As I mentioned earlier, the maximum capture area to be NIST compliant with the Canon EOS 30D (resolution of 3504 x 2335) is approximately 3.5 inches by 2.33 inches; for the 40D (with a resolution of 3888 x 2592), the area would be 3.8 inches by 2.6 inches. If you capture a larger area with those cameras, your image resolution would fall below 1000 PPI. Conversely, if you capture a smaller area, the resolution is going to be significantly higher. For example, if you capture an area with the Canon EOS 30D that is 1.75 inches by 1.16 (1 1/8), your resolution would be 2000 PPI, thus giving you a significantly better image to enhance should that be necessary. Once you have enhanced that image, however, your final and last step before submitting the image to AFIS / IAFIS would be to go to Image Size (in Photoshop), make sure the Resample Image checkbox is checked and the resample option is Bicubic or Bicubic Smoother, and then change the image resolution to 1000 PPI. (Please be sure to resave your adjusted image resolution file as a separate image so that you do not loose your high quality, high resolution image.)

The reality is that with a digital camera you always are restricted to the fixed resolution of the imaging sensor, and the PPI resolution will vary based upon how large or small of an area you try to capture with that camera. The major image quality issue for you is how clear the image appears when you view it on the screen (and at what zoom level) and what type of enhancement is required for the image. With a scanner (based on today's technology), you always will have a significantly higher resolution that you are required to have for NIST, and you should capture the image using the capture device capabilities (hardware ¦ also called optical ¦ resolution). In this case, the major determining factor is how large of an area are you going to capture. If you capture too large of an area with too high of an image resolution, your image file size with be gigantic.

In any event, your last step before saving the final processed image would be to set the resolution to 1000 PPI (the requisite NIST resolution) to accommodate the physical size requirement of AFIS/IAFIS hence, be compliant with the NIST requirement for the transfer of images between dissimilar systems. NIST never cared (much) about the resolution of the capture device; SWGIT added to the confusion by attempting to get people to always capture at the outer limits of the device (attempting to capture an area size based upon the number of photo diodes on the imaging sensor and meet the 1000 PPI) because most people did not (and many still do not) know how to calibrate (scale) images properly for 1:1 output when they had all different kinds of capture devices each with a different imaging sensor resolution. (Image quality is degreaded by capturing images based upon size of area and image sensor resolution! Plus this process further affects image quality because of the way digital cameras work -- the process did not consider the mosaic process used by all digital cameras for the determination of color values. That's another topic for another day!!!)

The only real capture requirement is that you must capture the image with the highest possible level of detail you still have to be the one who looks at the image and say yep, that image is clearly identifiable. If you can't say that, then you need to rethink who you are going to recapture the image to get something that you can use for identification (as well as processing).

I hope this clarifies the issue regarding the NIST requirement for 1000 PPI. As always, please do not hesitate to email me or call me if I can be of any further assistance.