The visual system transmits information from the retina to the brain along several different classes of parallel channel, each with its own characteristic responses. It is therefore inappropriate to describe visual responses in terms of a single overall modulation transfer function. We describe a psychophysical method for the measurement of spatial and temporal responses of single visual threshold mechanisms, present data for two such mechanisms and show how they can be used in the computation of responses to moving targets.

For most tasks of image evaluation, the human observer is very versatile and reliable. So he could be used as a good standard to measure the performance of technical image evaluation systems. However, to proceed in that way seems to be difficult since the human observer is open to many so-called subjective influences.

The optical design of biocular magnifiers at Pilkington PE Ltd is described with particular reference to the assessment of user visual comfort. The implementation of specialised computer software on a high speed computer dedicated to optical design calculation, has enabled accurate modelling of the visual image produced by very low relative aperture (f/0.6) magnifiers. Graphical output is discussed. The aberrations of the magnifier can be expressed in terms of visual effects and tolerances ascribed to define acceptable visual performance. The methods are applied to some new large diameter magnifiers for relaxed viewing.

The equivalent bar target technique for the tasks of detection, recognition and identification proposed by Johnson is used extensively to assess the performance of thermal imagers. However the approach does not enable background information to be incorporated in such an assessment. A new background criterion is therefore proposed which will enable backgrounds to be included in range modelling.

The operational performance of electrooptical imaging systems is currently predicted using models based on an approach introduced by Johnson in 1958. This model has been shown to be inadequate when used to predict performance over a wide range of conditions of signal-to-noise ratio and range to the target. This paper describes our effort to develop and validate a model which is more effective than the Johnson model in predicting image quality. The figure of merit is based on the hypothesis that image quality is related to the area of an aperiodic object which is barely detectable under the conditions of the test. This "minimum detectable target" approach is attractive in that not only is it an improved predictor of image quality, but also it can accommodate both noise limited and display contrast limited conditions and it can be based on a laboratory measurement of minimum detectable target size in the same manner that limiting resolution measurements are currently used with the Johnson criteria.

A unified computer model of image intensified sighting systems based on signal to noise performance modelling but incorporating the eyes varying properties as a function of luminance has been developed at the Royal Signals and Radar Establishment. This model has been shown to fit field data collected on military systems and has been in extensive use over the last 18 months at RSRE and within British Industry. This paper considers the validation of the model for image intensified night vision goggles and, by this example, the range of validity of the model will be illustrated.

Using standard Video techniques, a method and instrument have been developed for objective measurement of imaging performance, We expect the method to find application in various fields of imaging.

There exists a requirement to be able to measure the performance of a thermal imager objectively. A description is given of how the performance is measured subjectively at present, what the limitations are and describes how the subjective measurement may be made objective using a frame store and digital computer.

Measurements for quality assurance have to guarantee optical performance of production. Suitable criteria have to be in close connection with the application of the lens. First a simplified criterion to describe focusing with the human eye is given. Several performance criteria can be drived from MTF: MTF value for a characteristic spatial frequency. Integration of MTF value for a characteristic spatial frequency over the whole image field. In connection with this criterion the equality of image performance over the field has to be answered. Integration of MTF up to a limiting spatial frequency. This frequency can be specified by the intersection of MTF curve and threshold curve of the receiver. A classification of optical systems seems to be reasonable to yield characteristic parameters for simplification of performance criteria.

This paper describes a method for measuring and assessing the veiling glare of 35 mm camera lenses. It is a simplified version of our previous method which makes 54 measurements over the 24 mm X. 36 mm picture format, at all apertures [SPIE Proceedings 274-41 (1981), Opt. Eng. 21 343 (1982)1. In this case, measurements are made at only the following three positions in the 24 mm X 36 mm format, and at four apertures: along a diagonal, at the center and at ±0.9h, h being the maximum image height; at the maximum (full) aperture, the second maximum, medium, and minimum apertures. The criteria being that: if the maximum veiling glare value measured at the above three positions is not more than 1.5 %, it is very good; between 1.5 % and 3 %, good, provided that, in both cases, the ratio of the maximum to the minimum veiling glare value is not more than four; between 3 % and 6 %, poor, more than 6 %, very poor

The demands on a satisfactory image quality criterion are manifold: 1. to reduce the large number of data ensuing from an exact objective evaluation of an optical system, e.g. by OTF measurements; 2. to take into account specifications of the practical use of the optical system given by object selection and receiver characteristics or image processing units; 3. to be sensitive to variations of image quality which are relevant to the task of the imaging process, and insensitive to variations below a given level; 4. to be easily deduced from measurements and (or) theoretically calculated data. Such a criterion should be able to facilitate the comparison of different systems, to select the most suitable, to take into account the influence of quality variations due to fabrication tolerances, to help to construct the most adequate system by using the criterion as a merit function in automatic correction procedures and last but not least, to open up the possibility of optimizing an "adaptive optic" on line by the adjustment of variable parameters. It is demonstrated how for optical systems, Image Quality Criteria (I.Q.C.'s) can be deduced. Besides the selection of an ideal criterion, reaching agreement on a generally acceptable one and its use is of equal importance for its introduction.

A straightforward way to arrive at the spatial frequency content of an image is to digitize the image and then to compute the Fourier transform. This method involves expensive instrumentation and a powerful computer. If spatial frequency analysis is applied as a 100 % image assessment test in a production line it is neither suitable nor necessary to perform a complete frequency analysis. Commonly the spatial frequency response at a single well-chosen spatial frequency suffices to characterize the performance of the test specimen for adequate production control. By using fixed or tunable large-area spatial bandpass filters it is possible to build rigid and relatively inexpensive measuring instruments that can be used in a production line. For example, if the modulation transfer function has only to be measured at a single spatial frequency it is possible to use a checkerboard pattern which acts as a spatial triangular wave-filter on the image of the object line. If the line image is unsharp enough the higher harmonics of the filter have negligible influence. To arrive at this situation a well-defined amount of unsharpness can be arranged by the introduction of movement blurring. Using this concept we have built modulation-transfer-factor measuring instruments that are accurate to within two percent. The instruments described are used in the production lines for X-ray image intensifiers. This measurement replaces the measurement of the more subjective limiting resolution. A more detailed description of the method can be found in an earlier paper by the author (1). It is fairly easy to generate a tunable triangulare wave filter just by putting two bar patterns of equal frequency on top of each other with a small angle between the bars. The spatial frequency of the resulting triangular wave Moiree pattern varies with this angle. This principle is used in some rather expensive instruments for the measurement of the modulation transfer function. In these instruments the higher order harmonics of the triangular wave are filtered out by electronic means. Due to the electronic signal processing the tunable spatial filter acts effectively as a tunable spatial sinewave filter. There are other ways of producing tunable spatial sinewave filters. These and the application of such a filter in the measurement of the Wiener spectrum of photographs are the main subjects of this paper.

In this paper assessment of image quality is restricted to normal photography. That means that we should assess the capability of the lens to give a reproduction of the structures, the colours, and the intensity levels of any object in our surroundings, like landscapes, buildings, technical equipment, or people, animals, and so on. It means furthermore, that we do not look at an aerial image as in a telescope, but that always a photochemical process is included, leading to a projected, slide or to a printed image. The final image is inspected subjectively by the human eye. If we test a lens by measurement of data of an aerial image, the evaluation of the data should take the film, and the eye into account at least in a simplified. way. A second restriction is restriction to black. and white photography. Of course it is possible and desirable to incorporate the assessment of chromatic aberrations. But in this paper we will regard intensities only to keep problems simple.

A 50mm f/2 camera lens has been circulated to approximately 9 different laboratories throughout the world together with a specification carefully defining test conditions for a set of MTF measurements. The paper described the test specification and the results of the measurements obtained from the different laboratories. It also presented the results of a computer analysis with attempts to establish quantitively the cause of differences between sets of results. In particular the results have been analysed with regard to mechanical misalignment, differences in choice of focal planes, normalisation errors, spatial frequency etc.

Three MTF machines are compared: a lloir4 scan of spread function (ODETA V). With this method the Fourier transformation of the spread function is done instantaneously by the PM tube integrating over the image. The microprocessor tasks are limited to A/D conversion, scan control, normalising and data print out. b Slit scan of spread function. Using this method the spread function itself is loaded into the computer memory (N samples over the x axis). After data acquisition the computer starts an FFT algorithm to produce the MTF. c Knife edge scan of spread function. With this method an edge instead of a slit is used. The rest is the same, but the signal is the spatial integral of the spread function. Let A = total number of photons in spread function during the measuring time, and N = number of samples along x axis (for ODETA, N = 400).

A new approach to the numbering of detected interference fringes has been developed to enable computer analysis of interferograms containing closed-loop fringes, as captured by a TV-based lens testing Twyman-Green interferometer.3 The method requires some degree of operator intervention to set initial and maximum fringe numbers and information regarding the closed-loop fringes but otherwise is automatic. Following the fringe detection and numbering procedures the controlling computer evaluates "best-fit" values of the coefficients to predetermined aberration polynomials as used to represent the lens wavefront aberration presented in the interferogram. The wavefront aberration has been modelled by two different aberration polynomials, namely the Seidel and Zernike aberration polynomials. The latter polynomial includes additional terms to those in the former, enabling it to fit a wider range of asymmetries in the measured interferogram. It is therefore of considerable interest to compare the 'quality' of the fits obtained using the two aberration polynomials for a variety of interferograms. Results of the total process of data collection, polynomial fitting and lens Optical Transfer function evaluation are presented. The paper compares the residuals of fit, OTF curves and reconstructed interferograms derived from the same interferogram using the two polynomials. These suggest that, in general, the Zernike polynomials are more efficient than the Seidel polynomial.

The result of a measurement of the modulation transfer function is only useful as long as it is accompanied by a complete description of all relevant measuring conditions involved. For this reason it is necessary to file a full description of the relevant measuring conditions together with the results. In earlier times some of our results were rendered useless because some of the relevant measuring conditions were accidentally not written down and were forgotten. This was mainly due to the lack of concensus about which measuring conditions had to be filed together with the result of a measurement. One way to secure uniform and complete archiving of measuring conditions and results is to automate the data handling. An attendent advantage of automation of data handling is that it does away with the time-consuming correction of rough measuring results. The automation of the data handling was accomplished with rather cheap desktop computers, which were powerfull enough, however, to allow us to automate the measurement as well. After automation of the data handling we started with automatic collection of rough measurement data. Step by step we extended the automation by letting the desktop computer control more and more of the measuring set-up. At present the desktop computer controls all the electrical and most of the mechanical measuring conditions. Further it controls and reads the MTF measuring instrument. Focussing and orientation optimization can be fully automatic, semi-automatic or completely manual. MTF measuring results can be collected automatically but they can also be typed in by hand. Due to the automation we are able to implement proper archival of measuring results together with all necessary measuring conditions. The improved measuring efficiency made it possible to increase the number of routine measurements done in the same time period by an order of magnitude. To our surprise the measuring accuracy also improved by a factor of two. This was due to the much better reproducibility of the automatic optimization, which resulted in better reproducibility of the measurement result. Another advantage of the automation is that the programs that control the data handling and the automatic measurement are "user friendly". They guide the operator through the measuring procedure using information from earlier measurements of equivalent test specimens. This makes it possible to let routine measurements be done by much less skilled assistants. It also removes much of the tedious routine labour normally involved in MTF measurements. It can be concluded that automation of MTF measurements as described in the foregoing enhances the usefulness of MTF results as well as reducing the cost of MTF measurements.

With the increasing amount of computing power available per pound of money spent, the computer becomes an attractive tool to provide continued and improved quality control without increasing cost penalties, assuming that the tool has been used correctly and that we do not find ourselves using a sledge hammer to crack a nut. Equally when looking for effective ways of quality assurance it must not be considered that adding on a computer will automatically reduce the test cost or indeed improve the test/inspection results. A fundamental understanding, possibly gained during the develop-ment stage of the production equipment, of the performance parameter to be assessed is invaluable to the engineer. Progress is then made towards an optimisation of the time/cost of testing to ensure that the quality requirement of the product is met with the minimum of testing, initiating computer help at this stage further aiding test costs. Computer aid is also most useful in the development of products enabling many detailed results to be taken and accurately recorded in the minimum time, the elapsed time aspects are important in most development contracts as the customer is usually most anxious to receive the early development models, sometimes at the expense of testing. It is therefore required to make the most use of the time available.

The equipment is intended primarily for measuring the Optical Transfer Function (OTF), or Modulation Transfer Function (MTF), of lenses and afocal systems operating in the infrared (up to 14 m) or visible band of wavelengths. The OTF/MTF is obtained by measuring the line spread function (LSF) of the optical system and then using a microcomputer to convert this to the OTF or MTF. The microcomputer is also used to control motorized mechanical movements on the optical bench. However, the software has been written, so that where these are not provided the microcomputer will give the step-by-step instructions for doing the manual operations necessary to obtain the desired test results. The system will accommodate the use of optical bench facilities which are partly motorized and partly manual.

The influence of surface roughness on the optical performance of diamond turned components has been investigated in the wavelength range 7.3μm to 12.6,μm, with special regard to the Modulation Transfer Function (M.T.F.). Specifically, diamond turned, planar mirrors, and conventionally polished mirrors have been inserted in a selection of optical trains, in order to perform comparative measurements. The optical trains employed, corresponded to two commonly used assessment techniques, (i) Fizeau Interferometry, (ii) Measurement of the Line-Spread Function. It is known that surface roughness can give rise to a finite degradation in performance and may influence the M.T.F. It has been shown that method (i) is largely insensitive to such a degradation. However, method (ii) exhibits a significant dependence on surface roughness. Results are presented for mirrors constructed from aluminium alloy, and brass.

In solid state imaging arrays aliasing is most certain to occur, unless some sort of presampling filters are used. A presampling filter would be an optical low-pass filter. Using incoherent light, there is a limit to how ideal such a filter can be made. Loss of contrast will always occur during prefiltering. Therefore there will be a trade-off between aliasing and contrast loss. A review of this problem is given. It is argued that with some applications, aliasing is not very harmful. This corresponds well to the common, practical use of solid state imagers, where no anti-aliasing filters are applied.

The increasing use of solid-state array image sensors such as CCDs, CIDs and photodiode arrays, brings new aspects to the design and assessment of imaging systems. Some of the traditional image degradation mechanisms associated with tubes are still present, such as the temporal noise and limited resolution, and their effects on MTF and visibility. New problems include particularly the element-to-element non-uniformities in response and dark current which lead to structure on the picture and cause errors in systems with automatic image processing. These exhibit complex behaviour with temperature and wavelength, and give rise to complex calibration requirements. Many properties must be measured, some under varying environmental conditions, in order to fully characterise such arrays. The paper gives a broad description of the principal properties, their dependences on various conditions, and the testing methods employed, together with examples of typical results.

In the IR field 2D arrays of detectors are currently being developed for a large range of applications. In many cases it has been assumed that the "image" will be analysed by a computer rather than a human being. Such "automatic" devices may require totally different specifications from that required by the human operator. This presentation aimed at providing a visual illustration of a human observer with the 2D array image dealing in turn with: (a) Detector area fill factor (b) Minimum number of pixels (c) Shaping the pixel (d) Image uniformity (e) Noise characteristics While no specific specification was given, the relative importance of each parameter was discussed.

A flexible, computer-aided test facility designed to measure performance characteristics of IR-focal plane arrays is described. In addition, a concise set of performance descriptors is presented which describe the FPA characteristics in terms adapted to IR-system performance prediction and modelling.

Instruments for measuring the full aperture transmittance of thermal imaging lenses operating in the 8 to 14 micrometre spectral region are described, and examples of typical results obtained shown. The causes of discrepancies between the actual transmittances measured and those predicted from a combination of the anti-reflection coating performances expected and the material transmittances are discussed.

Optical images are invariably degraded by glare light and it has been known for many years that a substantial amount of glare can arise from multiple reflections. In fact, by far the principal source of glare from a polished, uncoated refracting surface is the reflection. The existence of these unwanted reflections led to the development of antireflecting coatings which have been very successful in suppressing the reflections, but such coatings nearly always introduce scatter. In making such coatings it is the usual practice to introduce plane-parallel witness pieces into the coating chamber and then to measure their reflectance, which is assumed to be typical of the reflectance of the batch. Investigations at Sira and elsewhere have shown that it is also important to control the scatter as this can have a considerable effect on the glare. However, many manufacturers state they are "not called upon" to measure scatter and specifications for antireflection coatings frequently make no mention of scatter. It is the purpose of this paper to suggest methods of testing coated optical components and witness pieces so as to produce readings which are closely related to the resulting glare in the image plane. Thus the relative importance of scatter and residual reflectance can be determined and coatings on components with curved surfaces can be assessed independently of the coating contractor, possibly after a period of use. Although this paper is mainly concerned with scatter and reflectance it is also relevant to discuss the assessment of localised defects such as digs and scratches. The assessment of such defects is usually undertaken by the manufacturer's optical inspectors. As inspectors get more skilled in detecting scratches, they also become more critical and it is known that many components are rejected for defects which can have no possible effect upon the performance. Much effort has been expended in recent years in producing graded standard defects with which to assess component faults by making visual comparisons but, in the opinion of this investigator, a functional test is much to be preferred.

A brief review of the influence of atmospheric transmittance on image quality is given. The problems of predicting infrared transmittance from the standard meteorological quantities is discussed and illustrated by examples.

An analysis is presented of noise and resolution limitations of a typical visible light imaging system in producing a digital image, and effects of the convolution of the image with the Marr-Hildreth edge operator, and the zero-crossing interpolation scheme, on the accuracy of edge location.

Some work is described involving the development and evaluation of methods for enhancing significant features in images used for non-invasive examination of the human body. The objective is to facilitate diagnosis by a human observer. For example, computer simulated coherent image processing is shown to be effective for correcting imperfect point spread function, and for filtering out noise due to the random emission of photons. Finally, an attempt to eliminate completely the human element in obtaining a diagnosis from gamma scans of the liver is described, which uses machine pattern recognition.

Digital image enhancement of microscope images encompasses both resolution enhancement, and image simplification and organisation by geometrical transformations. A variety of digital filtering techniques can be used to enhance the spatial resolution of digitised microscope images. When applied to images of banded chromosomes, it appears that of those tested, the maximum entropy method produces the most acceptable results, but its computational cost puts it at a great disadvantage.

A digital image processing system is described, which has been assembled for the routine diagnosis and correction of a wide variety of photographic problems encountered in police work. The major hardware components of this system are outlined, and then a range of useful enhancement and restoration techniques are described showing examples. These methods range from simple contrast manipulations to complicated restoration processes involving 2 dimensional large format deconvolutions.

Photographs are sometimes degraded by various defects which make it difficult or even impossible for a human observer to extract the relevant information from the picture. Examples of degradations are: low contrast, perspective distortions, blur due to defocusing , camera or object motion, and atmospheric turbulence. To improve the image quality in the presence of these degradations an interactive digital image enhancement and restoration system has been set up in hardware and software.

A short survey is given on work in progress concerned with quality of photographic images, MTF-data statistics of photographic lenses, and veiling-glare of camera lenses.