Advanced Air Force weapons systems require the real-time processing of data obtained from a multitude of sensors at data rates which exceed current capabilities. In addition, the advent of advanced sophisticated radars and other new information sensing devices for use by the tactical forces has created a requirement for an economical method of communicating large volumes of data in a rapid secure manner. These requirements necessitate the development of parallel processing techniques such as optical processing. Therefore, the Air Force is actively pursuing scientific research and exploratory development into both optical image and optical signal processing. This paper will describe the ongoing Air Force programs in these areas with special emphasis on spatial light modulators, holographic optical elements, non-linear optical processing, and optical correlators.

A new holographic implementation of a sampling technique permits straightforward representations of two dimensional space-variant optical systems. The set of sample transfer functions required for the representation is sequentially recorded on a single holographic plate by utilizing a diffuser to produce phase-coded reference beams. The phase coding operation then acts to suppress crosstalk between the stored holograms when they are played back simultaneously. Because this approach does not depend on volume effects in the recording medium in an essential way, the holograms can either be produced digitally or optically. Basic concepts and the results of experimental investigations are presented and discussed.

A variety of novel coherent optical data processing techniques and results appropriate for image processing and missile guidance are presented. Emphasis is placed on the practical problems of maintaining correlations in the presence of degradations between the input and reference function.

A matched filter optical processor is described that is compact, lightweight, and operates in real time. The use of lightweight reflection hologram lenses as the Fourier transforming elements allows for compact folded light paths. The holographic optical elements were designed and analyzed using a holographic optics ray-tracing computer program (HOAD). It was predicted that compact holographic optics can perform satisfactorily for input imagery of moderate space-bandwidth product; high performance can be achieved if longer focal lengths (1 meter) are allowed. A photoconductor-thermoplastic (PTP) device at the input plane of the processor provides for real-time data insertion. Data is scanned or imaged from an incoherent source onto the PTP device, which forms a phase recording that modulates the coherent illumination at the processor input. The PTP device operates in real-time and can be erased and reused many times. Compact optical processors of the type described are particularly well suited to applications such as space vehicle data processing or terrain-matching for terminal guidance.

Images of a simple object were digitally computed by a backward propagation algorithm from phase and intensity data measured over a large numerical aperture that was spaced from the object by about half the aperture width. Images were produced for distinct object orientations, and multiplying the image transparencies improved resolution.

In an underwater acoustic holographic imaging system, relative motion between the imaged objects and the recording plane may destroy the image. If the relative motion is known, the recorded hologram may be corrected so that the degraded image can be recovered. The system used in the study consisted of a linear scanning receiving array and a perpendicularly-oriented cylindrical insonifier. The signals detected by each receiving element are com-pared with electronic reference signals, thus generating in-phase and quadrature components. In this paper, uniform relative motion of translation between an object and the recording plane is computer simulated. Both horizontal and vertical motions are considered. An algo-rithm for image restoration and reconstruction is then applied to these data. Three-dimen-sional perspective plots are used to illustrate the results.

The transition from feasibility demonstration to a practical working system is never an easy step. This paper discusses some of the problems that will be encountered during the development of a practical interactive hybrid optical/digital image processing system. It focuses on the optical processor of a hypothetical application (IMADITOP) to illustrate a few of the important questions which will arise: What processing approach should be used? How fast will the processor really run? What will be the quality of the processed output? How complex must the system be? How much will it cost? How reliable will the system be? What risks are involved? What improvements in technology would help?

Commercially available OCR systems operate on high SNR text, use highly non-linear, serial processing techniques to discriminate between symbols, and achieve vanishingly small error rates. With degraded text, however, these discrimination techniques are no longer effective and an optimal OCR system must use classical signal detection techniques, namely, cross correlation of the text with stored symbols. Such a system using an optical correlator was first demonstrated in 1964 but has never been developed commercially, partially because no techniqe existed for the rapid generation and erasure of optical filters and because cross correlation yielded insufficient discrimination between symbols. The advent of real-time optical modulators such as the PROM may remove the first objection and we show in this paper that careful processing of the correlation data can remove the second. We review the theory (optimal detection with letter decoding based on language statistics) and present some examples.

A subtraction scheme capable of handling in real time two incoherently illuminated scenes and providing the sign of the subtracted information is described. The scheme utilizes two Hughes Liquid Crystal Light Valves, onto which the two images to be subtracted are projected. One valve is analyzed in between crossed polarizers, while the other one is in between parallel polarizers, both polarizers being implemented with a polarizing beam splitter and a quarter wave plate. The common output image plane onto which both images are superimposed, displays an intensity proportional to the difference between the two inputs. This intensity "rides" on a constant background intensity thus displaying the difference signal as well as its polarity. The read-in and read-out beams could be coherent as well as incoherent, the latter being more desirable due to its speckle-free image. Experimental results obtained with incoherent illumination will be shown.

Investigation of optical spatial filtering as a method for restoring fogged photographic imagery indicates that acceptable restorations are possible if the exposure pattern features a low contrast or a low fogging level. At higher contrast or fogging levels, the restoration procedure suffers significant error. The error, which is due primarily to the nonlinear nature of the photographic process, appears as a failure to reproduce the lightly exposed regions of the photographic transparency.

The parallel processing of digital data in two-dimensional page form using holographic techniques is discussed. The Boolean logic operations of AND, OR, EXCLUSIVE OR, and COMPLEMENT are shown to be possible in a single holographic cycle. All bits on one binary data page are operated on with the corresponding bits on a second page. Single-step processors using unchangeable recording materials have numerous applications. Multiple-step processors using continuously-changeable recording materials allow further computing power including the construction of the other Boolean logic operations and arithmetic operations. This type of optical processing may be particularly important for the construction of parallel associative processors.

Recent developments in noncoherent optical processing suggest that hybrid incoherent optical/electronic systems represent a viable alternative to coherent optical systems for diffraction-limited parallel processing of two-dimensional signals. Favorable characteristics of incoherent systems include greatly simplified input transducer requirements and a redundancy or multichannel nature that makes them resistent to the blemish noise typical of coherent processors. In this paper we discuss the basic kinds or classes of hybrid optical/electronic methods for bipolar processing of two-dimensional signals using two-pupil optical systems: direct subtraction and carrier, pupil interaction and non-pupil interaction. Tradeoffs and limitations are considered, and a new method for realizing a general bipolar pointspread function with positive-real pupil transparencies is presented.

Optics provide a powerful tool where real-time processing of signals using correlations or fourier transforms are used. For this reason the Navy has from time to time explored optical processors. Each time they have ultimately been set aside in favor of other techniques, for reasons given here. In presenting this historical overview no attempt has been made to be encyclopedic. The Navy has and still does support research and development in many aspects of optics. Rather, a few highlights from a systems applications program point of view are touched upon that cover the past two decades.

New applications of the use of coherent optical processing techniques, especially matched spatial filtering and input format control, in radar and sonar signal processing are reviewed. Emphasis is given to specific problems such as long coded waveforms of thousands of bits, processing of coded phased array and pulse burst radar waveforms, generation of the ambiguity function for use in radar waveform analysis, and the use of space variant optical processing as a novel approach to radar signal processing.

Signal Detection Theory models are developed which yield comparisons of optical and digital Fourier transform techniques in terms of detectability of power peaks obtained in their Fourier domains. A stochastic model is first given describing the quantization noise introduced by the finite register size in digital transform computers. The signal detection models are then developed describing the detectability of a transformed signal among this kind of noise, with models given for fixed-point and floating-point machines and for signal-known-exactly and signal-unknown detection problems. Signal Detection Theory provides a number of useful results here including the optimum detection statistic to be used, decision criteria for choosing cut-off points, the performance curve of the detector, and detection indexes which summarize detector performance. Once the digital transform computer and the type of detection has been specified, the resulting signal detectability can be used to specify requisite signal-to-noise ratios which the optical processor must achieve to obtain performance in optical Fourier domain equivalent to the digital processor. Alternately, the digital machine can be specified as to fixed-point or floating-point number representation, truncation or rounding of results, and register length to match the detection performance of a given optical processor. Results are given demonstrating these comparisons in which register length, array size, number-representation, and type of detection are the major independent variables.

This paper will discuss the capabilities of a new signal processing device consisting of a light-emitting diode (LED), a photographic mask, and an area-array charge-coupled device (CCD). Such a processor can perform a broad variety of useful one-dimensional operations including linear transformation (e.g., Fourier, Walsh, Hankel), multi-channel cross-correlation, filtering, and high-density read-only memory. Any desired window function can be designed into the mask and any desired amount of window overlap can be obtained by appropriate clocking of the CCD. Its strengths include high-speed, compact size, ruggedness, reliability, and potential low cost. However, as in other analog sampled-data systems, its accuracy is moderate (the equivalent of about 8 to 10 bits). The incorporation of a real-time programmable mask into this system expands its capabilities into the nonlinear and recursive filter-ing realms (in addition to programmable versions of the above-mentioned linear operations) at the expense of system size, complexity and cost. In many applications the numerical computation capability of such a processor far surpasses that of its conventional electronic digital counterparts.

Charge coupled devices and surface acoustic wave devices can provide alternative analog realizations of optical systems. The ability to compute analog Fourier transforms in real time with electronic filters provides an alternative to all-optical signal processing. Both one-dimensional and two-dimensional Fourier transforms can be realized electronically.

It is well known that a scattering medium between a transmitter and receiver in an optical link causes power degradation and loss of spatial coherence in the received field. The spatial coherence loss due to scattering can theoretically be partially recovered by proper receiver design and by invoking specific forms of focal plane processing techniques to aid in restoring the degraded field. This paper reports the results of studies devoted to determining most advantageous focal plane design, assuming statistical knowledge of the scattered field at the receiver. The resulting performance of such processors is also presented in terms of standard communication criteria (collected SNR and digital bit error probabilities). The study has application to optical space transmission through turbulent atmosphere and to optical communications over the air-ocean channel, both of which have been statistically modeled in earlier literature. The basis of the focal plane design is to perform an orthonormal expansion of the scattered field into statistically independent field components. The required processing in the focal plane is then dictated by the proper method for collecting the field components with respect to the desired processing criterion. A result of the study, applied to spatially homogenious scattered fields, is the development of a ring type of photo detector array properly spaced within the receiver focal plane. Performance of such devices is presented.

A new technology must continue to grow and prosper if its continued existence is to be insured. Although the field of optics has existed for many years, the extension of optics into viable electro-optic signal processing is relatively new. The growing pains of any new technology are severe and those of electro-optics are particularly stressing because of the apparent lack of direction and confused terminology. These issues are discussed and several solutions offered to spur additional creative thought in the hope of establishing realistic technology goals and guidelines.

A highly specialized optical processing system for generating ambiguity surfaces is described. The surfaces are generated by a series of cross-correlations between two raster-recorded, large time-bandwidth signals, one of which has been doppler compensated. High data throughput is achieved by performing nearly all necessary operations in parallel, optically. After initial detection, the signals are stored in raster format on an optical data buffer. Doppler compensation is effected by a newly designed, incoherent, Optical Doppler Transformation system which provides a controlled, variable anamorphic compression or expansion of the entire signal raster, and cross correlations are performed by a coherent optical correlator which has real time, optically addressed, spatial light modulators in the input and filter planes. This paper discusses the operation of the components of this ambiguity function generator, with particular emphasis on the new Optical Doppler Transformation system, and presents throughput performance predictions.

A real-time, no-moving-parts, tracking receiver that operates in the infrared region of the spectrum (λ = 905 nm) has been constructed. It optically processes the signal by means of a specially constructed holographic optical element. In one sense, this element can be thought of as an exotic optical lens with two, or four, off-axis foci. Actually, it is more than that and because of its additional holographic characteristics, it can be used for signal processing. This optical signal processing is based on the fact that the amount of light diffracted to each of the two, or four, off-axis foci is dependent on the physical position that light passes through the element. Thus, when a properly made holographic element is located at the field-lens position with detectors placed at the two, or four, image points, a real-time elec-trical tracking signal results. A number of these holographic optical elements have been fabricated, tested, and in-corporated into a tracking receiver that can track a GaAs laser illuminated target.

An optical channel is two-dimensional and, with thousands of elements in each spatial dimension, can provide a time-bandwidth product in the millions. The major limitation to the practical utilization of this capability, for real-time optical computing at high data rates, had been the input transducer. The development of the General Electric Coherent Light Valve (CLV) provided the required video input transducer, but the system bandwidth was restricted to the 20 MHz region. Continuing development has extended the CLV system signal bandwidth; recent experiments have shown a useful frequency range in excess of 100 MHz with a BT of 105. This paper describes the considerations involved in extending the capability of a CLV optical processor to permit processing gigahertz bandwidth signals.

This paper documents work performed at the McDonnell Douglas Astronautics Company in the field of optical signal processing as related to the processing of radar ambiguity functions. Included is a brief explanation of the two-dimensional range/doppler ambiguity function and a scheme for optically implementing it. Data is simulated digitally and optically. The optical process is documented at each point within the system. The final optical result is compared with the digital processor result, showing parallels in the fidelity of the ambiguity function.

Acousto-optic cells have found a major application in providing compact real-time electrical to optical transducers for optical signal processing systems(1). Such acousto-optic devices with bandwidths up to 1.0 GHz(2) and time bandwidth products exceeding 1000 may be configured with other optical components and devices to perform signal spectrum analysis, correlation, and signal sorting. This paper will present a conceptual overview of the principal acousto-optic techniques available for processing of wideband electronic signals.