In the last decade, low light level intensifier systems have come into widespread use not only in the military but also in such areas as law enforcement, medicine, and spectroscopy. The cascaded first generation intensifier has been produced in large quantity and has been commercially available for quite some time. The second generation intensifier has been declassified as of 1971 and is still in the developmental stage. The necessary gain is obtained by electron multiplication in the Multi-Channel-Multiplier (MCP). Con-sisting of only one stage, the second generation image intensifier has the advantage of smaller size and less weight. Figure 1 shows a 25mm first generation, a 25mm electrostatically focused and a 25mm proximity focused second generatioq image intensifier (inverter). The principle of operation for these types of image intensifier tubes will be described shortly. A more detailed discussion of the various image degrading processes will be given in the discussion.

The performance of infrared television camera tubes and infrared scanning systems are compared as a function of background photon flux density. The two-dimensional vidicon in the return beam mode is found to be a highly sensitive device for low background levels as encountered in space. At higher background levels, for application to terrestrial scenes, the two-dimensional vidicon is generally less sensitive than a linear array operating in the 8-l4μm region if vidicon nonuniformities exceed about 0.05%. It appears that the uniformity requirement for infrared television camera tubes in ground based operation is likely to be the limiting factor.

The modulation transfer functions (MTFs) of many optical and electro-optical imaging devices have the form (Ref. 1); T(f) = exp -(f/fc)n (1) where T(f) is the MTF, f is the spatial frequency in cycles/mm, fc is the frequency constant in cycles/mm, and n is the MTF index. It is usually possible to find values for the two MTF parameters fc and n such that the measured MTF agrees with equation (1) to within +0.05. Electro-optical devices have been classified (Ref. 2) in terms of the MTF parameters associated with each device. Now it is found that it is possible to extend this method of classification to optical devices and image transfer/display systems. In addition, the edge-response functions, line-spread functions, and point-spread functions associated with MTFs of this form are described, and the transformation equations are given.

Present methods being used for the evaluation of the peak-to-peak signal and RMS noise are perhaps satisfactory at high signal-to-noise levels, but at low signal-to-noise levels, where many low light level tubes operate, these methods are inadequate. Even at large light levels, signals which can ultimately be detected by the human eye (which integrates for periods of up to 1/5 second) will be buried in the noise of the video waveform (which is usually determined by the large bandwidth of about 10 MHz in the video channel). Any routine image tube measurement, such as spectral response or spatial resolution, will therefore need an averaging technique of one kind or another to eliminate the video noise and permit an accurate measurement over the entire useful range of signal levels. In addition, certain applications also require a knowledge of the noise itself. For instance, the amplitude distribution of the noise is very important in determining the "false alarm rate" of a system designed to detect a specific target. At the limiting resolution of the system, a large noise pulse would be indistinguishable from the target itself. That noise pulse would be a "false alarm". The false alarm rate of a system can only be predicted if the amplitude distribution of the noise is known. In the analysis of photoelectronic imaging devices, the present methods for measuring signal and noise (Ref. 1) include those where a time exposure of an oscilloscope trace is taken, allowing the film to average out the noise. The width of the "grass" displayed (assumed to indicate the ± 3a value of a Gaussian distribution) is divided by six and used as an estimate of the RMS noise value. Peak-to-peak signal is determined by measuring the positive extremes of the noise envelope. An improvement on this, while still subjective in nature, involves using a split screen presentation on the oscilloscope with the unknown signal on one side and a known RMS value on the other. By varying the known waveform amplitude until a match between the two sections is achieved, and then reading the value from the known source, the RMS value can be determined.

The optical transfer function of an optical component such as a lens or image tube is of great importance because this one function, together with any scene entering the lens, can be used to predict the scene leaving the lens. The concept of the optical transfer function, which is the product of a phase transfer function and a modulation transfer function (MTF) has been well described in the literature (Ref. 1). The analogous situation in electronics is that of an arbitrary waveform entering a linear electronic black box. Armed only with the transfer function, the engineer can predict the output waveform in the following way: the input waveform is decomposed into its Fourier components, each component is multi-plied by the value of the transfer function at that frequency, and the result is summed over all frequencies. The transfer function changes the amplitude and shifts the phase of each Fourier component.

The real time simulator ARTS), built by The Boeing Company to support in-house research and contract work, provides the ability to study individual sub-systems or end-to-end system performance. This capability includes sensors, optics, vehicle motion, data encoding, 300 MO quadriphase communications channel, ground processing operations, near real-time quick-look display, and hard copy image output.

The Return Beam Vidicon (RBV) is a high per-formance electronic sensor and electrical storage component. It can be operated with steady state or discrete exposures. At 100 1p/mm, RBV performance matches or exceeds the performance of high resolution film, particularly under low contrast conditions. Information can be read out with a single scan or in a multiframe manner for signal processing and for display on a CRT monitor. Electronic zoom can be employed effectively for data compression and for image magnification. The high performance and flexibility of the RBV permit wide application in systems for image sensing, storage, processing and transmission. This paper summarizes the principal characteristics of the RBV and describes recent improvements in performance achieved with a new electron gun of approximately 20 times the beam current capability previously available. These improvements include substantial increases in sensitivity and frame rate. Projections of performance are made for silicon target versions of the RBV currently under development for both sensing and electrical storage functions.

For many years image tubes have been made which use semitransparent photocathodes to conveniently separate the optical and electro-optical functions. It is now possible to employ opaque photocathodes in image tubes having a special electromagnetic lens, without the use of special internal image forming optical lenses or mirrors. The design of this special electron-lens and measurements of some of its characteristics are discussed in succeeding sections.

The development of the electro-optical image intensifier tube has created a need for new lens objectives uniquely designed for low-light-level applications,. Such lenses frequently resemble those intended for aerial reconnaissance purposes, but they have certain important design features that make them dif-ferent from their photographic counterparts.

Passive low-light level television has been used extensively for years in both mili-tary and civilian surveillance applications. The widespread usage of LLLTV comes from its capability of allowing the monitoring of areas which are necessarily dark by direct visual observation. Thus, the television surveillance system extends the observer's keenest decision-making sense -- sight. However, just as the human eye becomes limited at night, the passive LLLTV becomes limited on moonless, overcast nights and in poor visibility weather. To overcome this limitation and to increase the average usable range, an artificial covert illuminator is added to the television system. The illuminator not only alleviates the dependency of performance on naturally occurring nocturnal sources, but increases the system's capability to perform under poor visibility conditions.

Imaging techniques at millimeter wave-lengths have been developed using an image dissection panel. ^1-4 This work is an extension of the pioneering effRrts of Lasser et al in the infrared region, although a complete transformation of the image dissection tube to a solid state counterpart has taken place. Figure 1 shows our basic imaging concept in block diagram form. Various configurations of imaging systems have been explored including transmission and reflection modes, simple and compound lens ar-rangaments, and single and multiple collectors.

The purpose of this paper is to describe the image processing research at MIT and to present some qualitative considerations about picture quality.

The processing of images on a digital computer is usually done for the purpose of extracting information from an image which the unaided human visual system cannot extract by itself. In nearly all image processing procedures a large amount of a priori knowledge related to the image must be utilized in order to extract the desired information. By a priori knowledge we mean knowledge about the image obtained from sources other than the image itself. This knowledge might include information about the object, the transmission media, the optical system, the sensor, the scanning and digitizing equipment, and the processing technique. Often a part of the a priori knowledge is contained in the mind of the observer as a result of accumulated experience.

The Mariner 9 spacecraft was inserted into orbit around Mars on 14 November 1972. The spacecraft carried two slow scan vidicon camera systems; the wide angle camera had a focal length of 50 millimeters and an 11 by 14 degree rectangular field of view; the narrow angle camera had a 500 millimeter focal length and a 1.1 by 1.4 degree field of view. The spacecraft orbit had a 65 degree inclination and the orbital period was approximately twelve hours. During the first 120 days of the mission, images were taken on opposite sides of the planet within each 24 hour period. At the start of orbital operations, approximately 60 images per day were returned. The number of frames per day decreased toward the end of the mission, and a total of 7,000 frames were transmitted by June 1972. The spacecraft operations for a typical Martial orbit are shown in figure 1.

The speed of the holographic image deblurring method(Ref.1-4) has recently been further enhanced by a new speed in the realization of the powerful holographic image-deblurring filter. The filter permits one to carry out the deblurring, in the optical computer used,in times of the order of 1 second, in contrast to times of minutes ,and sometimes hours ,which would be required for the enhancement of similar photographs with methods using microdensitometer scanning,encoding and digital electronic computing, even with the fastest and most powerful electronic computers. The newly developed method for rapid realization of the 'extended-range high-resolution holographic image-de-blurring filter'(Ref.5)permits one to make the filter,by means of the photo-holographic method used,in less than a single day's work,starting from the point-spread function(impulse response function),rather than in months required in earlier forms. Several filter realization methods have been developed(Ref.1-7),one of which is made possible by a suitable combination of digital electronic computation of the 'impulse response function' [for cases when it may be derived by analysis ]and its 'photo-holographic'realization in the form of the 'holographic Fourier-transform division filter'(Ref.1,2),In this way we take advantage of the very fast optical computing for the deblurring(compared to much longer digital electronic computing times), while using the digital computer for the part of the filter computation in which it is very rapid. The experimental achievements using the holographic image-enhacement method will be illustrated with examples ranging from out-of-focus or motion-blurred photographs, including 'amateur'photosrecorded on POLAROID film, to the shar-pening of the best available electron micrographs of viruses, obtained from the most powerful scanning electron microscope(Ref.8),In one case, the method has permitted us(Ref.9)to resolve a long-standing biophysical controversy (Ref.10) by revealing the double-helical coiling of a filamentous bacterio-phage(fd) virus specific to E.coli (Ref.11) and by showing that only a portion [rather than perhaps the entire virus, as previously assumed according to some studies(Ref.12)]did in fact seem to take on the double-stranded structure. Images recorded with X-rays, notably from rocket-borne photos of the Sun, and out-of-focus photographs from ca-meras in NASA satellites have been similarly deblurred(Ref.13), and other among the many applications include the sharpening of images in ultrasonic imaging, beyond the ultimate as previously limited by the instrumental limitations themselves.

In the computer processing of dynamic images (cinefilm, TV) data volume and rates are usually excessively high. This is because of the addition of two spatial demensions to the normal time-domain processing problem. It has therefore been necessary to reduce the spatial and intensity resolution of the converted images. This paper examines the errors resulting from these quantization trade-offs, as applied to several medical image processing problems.

Many specific diagnoses of diseases of the heart can be obtained in cardiac catheterization laboratories. Dynamic changes in volume and distributions of coronary blood flow are studied by means of injecting roentgen-opaque (i.e., x-ray dense) material into the left ventricle or the right ventricle through a catheter whose tip has been threaded into the heart via a puncture in a peripheral vein or artery such as femoral artery in the leg. At the time of the injection of contrast media, the chest of the patient is irradiated with x-ray and dynamic movements of the opacified heart are projected onto a'fluoro scopic screen and then recorded on videotape by a television system or on film by a photographic system. Although cine roentgenography (that using film) can have higher spatial and temporal resolution (Ref. 1) than video roentgenography (that using television), the cine systems require a longer time to complete the diagnosis since the film must be developed.

Importance of Magnetic Fields In 1908, just 12 years after Zeeman discovered that magnetic fields cause a polarization and shift in certain spectral lines, Hale used this effect to detect the presence of strong magnetic fields in sun-spots. Further studies have shown that magnetic fields exist over the entire surface of the sun in very complicated patterns and determine the structures and motions of the solar plasma such as prominences, spicules, and flares. The very strongest fields of several thousand gauss cause the spectral lines to actually split but for weaker fields the Zeeman shift is only a small fraction of the line width and to measure this shift four steps are required: spectral dispersion or isolation of a magnetically sensitive line, selection of one state of circular polariza-tion, detection of the light intensity, and finally the subtraction of the light in the two opposite states of polarization. Spectral isolation can be achieved by a spectrograph or optical filter; the polarization states can be selected by an electro optic crystal or quarter wave plate with polarizer; the detector can be a phototube, film, or television camera; and the subtraction can be electronic, photographic, or digital. This paper describes a new type of solar magnetograph using digital techniques.

The 3M Brand EBR-700 series Electron. Beam Image Recorder has been developed to produce corrected 50mm images on 70mm film of data from the camera systems aboard the ERTS A spacecraft. The images are produced by raster scanning with direct electron beam bombardment on film. The machines are capable of producing high quality 70mm black and white film from monochrome video signals. The resolution, (in excess of 6000 TV lines), dynamic range (in excess of 100 to 1), geometric accuracy,. and sensiometric fidelity provide a near-transparent window through which the outputs of the Return Beam Vidicon and Multispectral Scanner camera systems can be displayed.

Spacecraft imaging systems produce amounts of data which can easily exceed the down-link channel capacity. This paper demonstrates data compression methods which reduce significantly the communication requirements with minimum loss of information in the received imagery. Differential PCM data compression takes advantage of the lower entropy of the scene-invariant brightness difference signal. Entropy coding of the compressor output provides maximum signal-to-noise ratio per channel bit. Two-bit-per-sample imagery is shown to be equivalent to 6-bit PCM. Reconstruction error and subjective evaluation of actual images are used to measure compression performance.

An Image Converter (IC) has been developed for image chain analysis of present and future imaging systems. The IC possesses extreme flexibility for simulation of operational variables, and is used in conjunction with computer facilities for processing of the video data. A CRT, laser, or solid state array is used in the read mode, and a CRT or laser in the print mode. The IC has been used to study image data coding, data compression and transmission, utilizing its ability to print the original data, reconstructed data, and/or display the error data in spatial form. It has been used to investigate sample size, SNR, attitude control errors and bit error rate effects on PCM and differential PCMts and inverse MTF.

The image forming ability of an IRLS system has been assessed by means of laboratory measurements and in-flight testing using a ground test array. The laboratory tests have included determination of the over-all system modulation transfer function, amplitude characteristics and the minimum resolvable temperature. The ground test array consists of a 3-bar wedge pattern and a radiance step constructed of dissimilar panels of painted and sand-blasted aluminum. Analysis of the imagery resulting from over-flying the array yields data of the amplitude characteristics and the noise-limited resolution of the system.