Издательство John Wiley, 1991, -180 pp.
The roots of optical signal processing date back to the work of Fresnel and Fraunhofer nearly 200 years ago. But the connection between optics and information theory did not take shape until the 1950's. In 1953, Norbert Wiener published a paper in the Joual of the Optical Society of America entitled "Optics and the Theory of Stochastic Processes" A). That same issue contained articles by Elias on "Optics and Communication Theory" B), and by Fellgett on "Conceing Photographic Grain, Signal-to-Noise Ratio, and Information" C). Other interesting papers of that decade include those written by Linfoot on "Information Theory and Optical Imagery" D), by Toraldo on "The Capacity of Optical Channels in the Presence of Noise" E), and the seminal paper by O'Neill on "Spatial Filtering in Optics" F). These papers represent the early infusion of information theory into classical optics.
A powerful feature of a coherently illuminated optical system is that the Fourier transform of a signal exists in space. As a result, we can imple- ment filtering operations directly in the Fourier domain. This feature was anticipated in papers by Fresnel, Fraunhofer, and Kirchhoff, and had been demonstrated, before the tu of this century, by Abbe in connection with his work on images produced by microscopes.
It is one thing, of course, to recognize that images can be changed by modifying their spectral content; it is another matter to implement the change. In their image-processing work, Marechal G) and O'Neill F) used elementary spatial filters to illustrate the principles of optical spatial filtering and to perform mathematical operations such as differentiation and integration. Such was the status of optical signal processing in the early 1960's.
A major impetus to optical signal processing was the need to process data generated by synthetic aperture radar systems. These radar systems were a significant departure from conventional ones because they proved that a small antenna, when used appropriately, provides better resolution than that achieved by a large one. This result, at first glance, is surprising. No physical principles are violated, however, because the small antenna samples and stores the radar retus as a means to synthesize a large antenna. To display the radar maps, we need to process the two dimen- sionally formatted radar retus; because digital computers could not handle the computational load, powerful new signal-processing tools were required.
Photographic film stored the extensive information collected by the radar system. Range information was stored across the film and azimuth information was stored along the film. When the film was illuminated with coherent light, the desired radar map was created by the propagation of light through free space, coupled with the use of some special lenses (see Chapter 5, Section 5.6, for more details). Generating radar maps was the first routine use of optical processing and was the first application for which the matched spatial filter included complicated phase functions such as lenses. It is hard to overestimate the influence that radar processing had on optical signal processing and holography. The classic paper by Cutrona, Leith, Palermo, and Porcello on "Optical Data Processing and Filtering Systems" (8) is important because it presented the basic concepts in a remarkably complete way.
To expand the capabilities of optical filtering to more general opera- tions, such as matched filtering for patte recognition, we needed to construct filters for which amplitude and phase responses were arbitrary. A solution to the difficult problem of recording the phase information was developed in the early 1960's (9). Because every sample of an input object contributes light to every sample in the matched filter, these two planes are globally interconnected. The computational power of such systems is high because many complex multiplications and additions are performed in parallel. The performance of patte-recognition systems from that decade has yet to be exceeded.
Basic Signal Parameters
Geometrical Optics
Physical Optics
Spectrum Analysis
Spatial Filtering
Spatial Filtering Systems
Acousto-Optic Devices
Acousto-Optic Power Spectrum Analyzers
Heterodyne Systems
Heterodyne Spectrum Analysis
Decimated Arrays and Cross-Spectrum Analysis
The Heterodyne Transform and Signal Excision
Space-Integrating Correlators
Time-Integrating Systems
Two-Dimensional Processing
The roots of optical signal processing date back to the work of Fresnel and Fraunhofer nearly 200 years ago. But the connection between optics and information theory did not take shape until the 1950's. In 1953, Norbert Wiener published a paper in the Joual of the Optical Society of America entitled "Optics and the Theory of Stochastic Processes" A). That same issue contained articles by Elias on "Optics and Communication Theory" B), and by Fellgett on "Conceing Photographic Grain, Signal-to-Noise Ratio, and Information" C). Other interesting papers of that decade include those written by Linfoot on "Information Theory and Optical Imagery" D), by Toraldo on "The Capacity of Optical Channels in the Presence of Noise" E), and the seminal paper by O'Neill on "Spatial Filtering in Optics" F). These papers represent the early infusion of information theory into classical optics.
A powerful feature of a coherently illuminated optical system is that the Fourier transform of a signal exists in space. As a result, we can imple- ment filtering operations directly in the Fourier domain. This feature was anticipated in papers by Fresnel, Fraunhofer, and Kirchhoff, and had been demonstrated, before the tu of this century, by Abbe in connection with his work on images produced by microscopes.
It is one thing, of course, to recognize that images can be changed by modifying their spectral content; it is another matter to implement the change. In their image-processing work, Marechal G) and O'Neill F) used elementary spatial filters to illustrate the principles of optical spatial filtering and to perform mathematical operations such as differentiation and integration. Such was the status of optical signal processing in the early 1960's.
A major impetus to optical signal processing was the need to process data generated by synthetic aperture radar systems. These radar systems were a significant departure from conventional ones because they proved that a small antenna, when used appropriately, provides better resolution than that achieved by a large one. This result, at first glance, is surprising. No physical principles are violated, however, because the small antenna samples and stores the radar retus as a means to synthesize a large antenna. To display the radar maps, we need to process the two dimen- sionally formatted radar retus; because digital computers could not handle the computational load, powerful new signal-processing tools were required.
Photographic film stored the extensive information collected by the radar system. Range information was stored across the film and azimuth information was stored along the film. When the film was illuminated with coherent light, the desired radar map was created by the propagation of light through free space, coupled with the use of some special lenses (see Chapter 5, Section 5.6, for more details). Generating radar maps was the first routine use of optical processing and was the first application for which the matched spatial filter included complicated phase functions such as lenses. It is hard to overestimate the influence that radar processing had on optical signal processing and holography. The classic paper by Cutrona, Leith, Palermo, and Porcello on "Optical Data Processing and Filtering Systems" (8) is important because it presented the basic concepts in a remarkably complete way.
To expand the capabilities of optical filtering to more general opera- tions, such as matched filtering for patte recognition, we needed to construct filters for which amplitude and phase responses were arbitrary. A solution to the difficult problem of recording the phase information was developed in the early 1960's (9). Because every sample of an input object contributes light to every sample in the matched filter, these two planes are globally interconnected. The computational power of such systems is high because many complex multiplications and additions are performed in parallel. The performance of patte-recognition systems from that decade has yet to be exceeded.
Basic Signal Parameters
Geometrical Optics
Physical Optics
Spectrum Analysis
Spatial Filtering
Spatial Filtering Systems
Acousto-Optic Devices
Acousto-Optic Power Spectrum Analyzers
Heterodyne Systems
Heterodyne Spectrum Analysis
Decimated Arrays and Cross-Spectrum Analysis
The Heterodyne Transform and Signal Excision
Space-Integrating Correlators
Time-Integrating Systems
Two-Dimensional Processing