Stephen D. Fantone, David A. Imrie, Daniel Orband, and Jian Zhang
Wakefield, MA 01880 USA
The introduction of third generation thermal imagers brings a new challenge to the laboratory evaluation of the thermal imager resolution performance. Traditionally, the Modulation Transfer Function (MTF) is used to characterize the resolution performance of the thermal imager. These new third generation of thermal imagers can be categorized as sampled imaging system due to the finite pixel size of the elements comprising the focal plane array. As such, they violate the requirement of shift invariance required in most linear systems analyses.
We present a number of approaches to measuring the resolution performance of these systems and conclude that source scanning at the object plane is essential for proper MTF testing of these sampled thermal-imaging systems. Source scanning serves dual purposes. It over-samples the intensity distribution to form an appropriate LSF and also generates the necessary phases between the thermal target image and the corresponding sensor pixels for accurate MTF calculation. We developed five MTF measurement algorithms to test both analog and digital video outputs of sampled imaging systems. The five algorithms are the Min/Max, Full Scan, Point Scan, Combo Scan, and Sloping Slit methods and they have all been implemented in a commercially available product.
Advanced thermal imaging systems now incorporate two-dimensional staring imaging arrays. The elimination of mechanical scanning results in a more reliable and higher sensitivity imaging system. These detector arrays essentially sample the image presented to them by the imaging optics. The detector signal is then digitized and the camera system usually produces a two-dimensional digital image.
While the lens that images the object onto the detector can be readily characterized utilizing traditional Modulation Transfer Function (MTF) measurement equipment, the detector array violates the requirement for linear systems that each component in the system must be linear and space invariant. This is due to the sampling nature of the imaging array and the finite size of each pixel.
In this paper we discuss our approach to characterizing detector array performance.
2. SYSTEM TESTING
Figure 1 depicts the major components and testing points for a typical thermal imaging system.
In a typical test configuration, a thermal target is presented to the imaging system that is composed of a optical system (lens), a detector, a preamp, electronic processing and a display system. This system can be interrogated at a number of test points to assess the performance of either an individual component, e.g. lens, detector, electronics or display, or the system in its entirety.
Reference to Test points for OpTest refer to use of Optikos’s OpTest1 product line which is a family of metrology equipment intended for evaluating the quality of images formed by an optical system. Test points for I-SITE refer to use of Optikos’s I-SITE2 product line which is intended to evaluate the signal produced along the signal processing chain, e.g, detector output, the digital or analog video signal itself, or the output of a reference camera (Photometric Camera) that can evaluate the image produced on the display itself.
In earlier generations of thermal imagers, a single element detector with a 2-dimensional scanner was used to create a full two dimensional image. This was followed by the use of linear detector arrays that scanned subfields of the scene and eventually line scans of the entire scene. In recent years, full two-dimensional thermal imaging staring arrays have become economical, and have generated an associated need to properly characterize their imaging performance.
Figure 2 – Performance metrics that are associated with system level testing.
The primary measures of system level performance are the minimum resolvable temperature difference (MRTD) and minimum detectable temperature difference (MDTD). Obviously, there are many other performance metrics that must be satisfied as well and they are usually considered of secondary importance.
When using sampled devices such as staring sensor arrays, assessment of the resolution of the system is more complicated due to the lack of shift invariance in the system. In conventional linear systems, an incremental shift in the input of a system in time or space should translate directly in the output to a corresponding shift in time and space. Pixelated devices violate the assumption of space invariance since, in most cases, shifts within the size of the pixel do not produce corresponding shifts in the output. This leads to questions as to what the appropriate metric for staring array sensors should be. Depending of the alignment of the test target (a slit) to the detector array, substantially different measurements MTF measurements can be obtained as is illustrated in Figure 3.
Figure 3 – The principle of stationarity and shift invariance is not valid for sampled and pixelated systems. Note that the output of the staring array is dramatically different depending on the alignment of the slit object relative to a column of pixels.
3. METHODS OF MEASUREMENT
We utilizes five different approaches to characterizing the MTF of staring array systems.
• MTF Min/Max Method
• Full Scan Method
• Point Scan Method
• Combo Scan Method
• Sloping Slit Over-scanning Method
We shall consider each of these in turn. The first four incorporate source scanning, that is, the object itself is moved to achieve movement of the image in the detector plane. If source scanning means are not available, the Sloping Slit method provides an alternative approach.
3.1 MTF Min/Max Method
In this case we locate a slit target at two different locations relative to a row or column in the detector array. We calculate the MTF of the slit image. By using the source scanning drive, we adjust the location of the slit image to achieve the highest MTF and record their values. We then adjust the position of the slit image to the alignment position that provides the lowest MTF. Each of these MTFs are recorded. We then combine them as an arithmetic or RMS average to report the average MTF or sum MTF.
3.2 Full Scan Method
The Full Scan Method involves positioning the image of the slit object at different positions relative to the detector array so that the phase of the slit image can be adjusted relative to a column or row of the staring array. At each position, we record the line-spread function. We then record these line-spread functions as the image of the slit scans across the detector array. We shift the scanned LSFs an amount corresponding to the shift that would occur if the system were shift invariant and average the LSFs. The MTF is calculated using this averaged line spread function. This approach requires an oversampling frame grabber and hence is usually used with analog cameras to provide a metric for overall system performance.