Software speeds analysis of imaging peripherals
To speed fingerprint identification, the FBI (Washington, DC) is acquiring an integrated, automated fingerprint-identification system that incorporates image scanners, monitors, and printers. To ensure the proper calibration of these peripherals, the FBI has defined a series of image-quality specifications. One of the scanner performance parameters is the spatial frequency response or modulation transfer function (MTF). Because the MTF provides a measure of the contrast reduction of an imaging device as a function of spatial frequency, it can be used to check image quality.
To compute the MTF of scanners, cameras, or printers used in the system, the FBI contracted Mitre Corp. (Bedford, MA) to develop a software package that uses sine-wave targets to compute a peripheral`s MTF and gray-scale linearity, as well as to detect any aliasing that may occur. MTF parameters are determined by analyzing sine-wave input targets obtained from Sine Patterns (Penfield, NY). Sine-wave pattern targets were chosen because their signal-to-noise ratios are relatively high compared with other target types. In addition, the presence of aliasing, caused by the imaging system, can be rapidly determined by analyzing sine-wave image targets.
For use by a scanner, the input sine-wave target is placed on either paper or film. For a printer, the input sine-wave target is input in digital format, and the printer produces a hardcopy output. In this case, output is scanned to create the digital image necessary for input to the computer.
Determining the MTF of an image scanner is performed in several stages. First, the program reorients the image to a standard orientation and computes the resolution and skew angle. Gray patches and sine patterns are then located, and the average gray level of each gray patch is compared to the known target reference. Linear regression analysis then computes the linearity of the device; the best-fit line is used to convert sine-wave maximum and minimum (peak and valley) gray-level values into equivalent values in target space, where their MTF is computed. The peripheral`s MTF is then determined by dividing the image modulation at a given frequency by the corresponding target modulation.
Discrete Fourier transforms along selected rows determine whether any aliasing is present. Finally, the program outputs image corner coordinates, resolution, skew angle, linear regression, scanner linearity, and MTF.
Originally written in C language on a Sun Microsystems workstation under SunOS 4.1.3 software, the completed program has been ported using Watcom C to run on IBM PCs under DOS, Windows 95, and Windows NT. Now in the public domain, the MTF program and documentation can be downloaded at www.mitre.org/research/mtf. For more information, contact Norman Nill at email@example.com.