`
`Ex. GOOG 1023
`
`EX. GOOG 1023
`
`
`
`
`
`

`

`A Reprint from the
`
`SPIE Volume 1 082
`
`Applications of Electronic Imaging
`
`17-19 January 1989
`Los Angeles, California
`
`Critical technologies for electronic still imaging systems
`
`Michael Kriss. Ken Parulski; David Lewis
`Eastman Kodak Company, Rochester, New York 14650
`
`©1989 by the Society of Photo-Optical Instrumentation Engineers
`Box 10, Bellingham, Washington 98227 USA. Telephone 206/676-3290
`
`Ex. GOOG 1023
`
`

`

`Critical technologies for electronic still imaging systems
`
`Michael Kriss, Ken Parulski, David Lewis
`Eastman Kodak Company, Rochester, New York 14650
`
`ABSTRACT
`
`Electronic still camera systems are now in the consumer market place. The hard copy image quality of these systems
`is poor in comparison with the ever improving photographic film systems. However, the rate at which solid state
`image sensor technology, signal processing technology, mass storage technology, and non-photographic hard copy
`technology are advancing indicates that these electronic still camera imaging systems will someday find a place
`alongside traditional photographic systems. The current and future status of these critical technologies is the
`subject of this paper.
`
`1. INTRODUCTION
`
`On January 6, 1839, the Academie des Science in Paris announced that Louis Jacques Mande Daguerre had
`"discovered a method to fix the images which were represented at the back of a camera obscura; ... ". 1 Since that
`eventful day photographic images have dominated how mankind has recorded history from world wars to family
`outings and documented new discoveries ranging from the exploration of the atom and the tombs of ancient Egypt to
`the natural habitats of the rain forests, jungles, and deserts around the world. Before World War II there were no
`serious challenges to the photographic method of image recording, but the commercial development of color
`television in the 1950's and the subsequent development of high quality magnetic recording and VLSI semi-conduc(cid:173)
`tor technology in the 1970's and 1980's has brought electronic image recording to the consumer in the form of home
`video systems. The new 8 mm video camcorders have replaced the Super 8 film systems as the choice for recording
`family events and travel. In the commercial area, Electronic News Gathering, ENG, has replaced 16 mm film for
`television news broadcasting. Attempts are being made to use High Definition Television Systems ( HDTV) as a
`replacement for film in the motion picture industry. While HDTV systems have not replaced film for motion picture
`production, the introduction of the BETA and VHS VCR systems and the Laser Disc systems have brought film
`originated movies into the homes of millions.
`
`During the same time span, conventional silver halide-based still photography has had strong, continuous gTowth.
`This growth has been spurred by improvements in film, cameras and ease of processing. Today a consumer can spend
`less than $100 for a high quality 35 mm camera with autofocus, automatic exposure control, automatic film advance,
`automatic film speed indexing, and built-in electronic flash. The resulting images are of very high quality. But while
`conventional photography continues to enjoy strong growth there is another electronic imaging system appearing on
`the horizon, one that may someday share the consumer market with the film-based systems of today. The electronic
`still camera, ESC, is a commercial reality today, and it and the technologies that make it possible are the subject of
`this paper.
`
`1.1
`
`Electronic still camera system concept
`
`Figure 1 shows a conceptual ESC system that could be assembled from currently available products. The system and
`camera is built around the Still Video Floppy, SVF, which records the image as an analog video signal. 2 Figure 2
`shows the original SVF standard along with the new High-Band standard. In both cases the camera records 50 single
`field images or 25 full frame images; a video frame is made up of two interlaced fields. In the case of the High-Band
`standard, the images are recorded using a higher carrier frequency thus providing more bandwidth for each scan line
`and yielding greater horizontal resolution; the vertical resolution remains the same- 242 lines for the field format
`and 484 lines for the frame format.
`
`The player/recorder converts the SVF analog signal into a form suitable for display on a conventional television set or
`monitor and also allows one to capture images from broadcast television or from VCR's and record them on the SVF
`disks. By using an image transceiver with a modem and public or private telephone communication systems one can
`send images anywhere in the world. Hard copy can be obtained from images stored on the SVF disks. The prints can
`be made from any number of print engines including thermal printers, electrophotographic printers, ink jet printers,
`and raster printers exposing conventional or instant photographic materials.
`
`SPIE Vol 1082 Applications of Electronic Imaging (1989) / 157
`
`Ex. GOOG 1023
`
`

`

`SVF Unit
`
`Rim
`
`Amplitude
`
`Still Video System
`HomeFilmto~ ~
`:I!!! - ~ • !
`6-;;=~=Eii.
`
`Electronic Camera /
`
`Still Video Player/Recorder
`:
`t ~~:C..ivor
`Floppy
`(single or multi-disk)
`Telephone~
`
`~·~~
`
`TV-Photo Lab System
`
`Hard Copy Device
`
`Image Transceiver
`
`FM Color difference
`Signal
`
`FM Luminance Signal, Y
`
`---4 Normal ~
`r-High Band -4
`
`Frequency
`Deviation
`
`Frequency
`Deviation
`
`6
`5
`4
`Frequency (MHz)
`f1; R-Y Color difference, 1.2 MHz Center frequency
`f2; B-Y Color difference, 1.3 MHz Center frequency
`
`7
`
`8
`
`9
`
`Figure 1. Conceptual diagram of a po~sible still
`video system based on the still video fldppy, SVF,
`standard_
`
`Figure 2. The image encoding standard for SVF
`systems.
`
`An additional feature of the system is that a scanner can be used to convert existing images on negatives, transparen(cid:173)
`cies, or paper into electronic signals for recording on the SVF disks. Image scanners/recorders can be installed either
`at photofinishers or in the home. Such a system provides complete flexibility to the consumer.
`
`Figure 3 provides a more detailed look at the important parts of a ESC system; the system shown is just an
`abstraction of an ESC system and does not represent any particular product. One key aspect of such a system is that
`film-based images that are scanned into it can make use of the same system hardware and software that is used to
`transform the electronically captured image into a final hard copy print, soft display, or transmitted image.
`
`In what follows, detailed discussions will be presented on the key ESC technologies: the solid state sensors that
`record the image, the in-camera signal processing that is required, the recording technology that stores the images,
`and the hard copy technology that produces prints. In far less detail, the technologies that deal with data compres(cid:173)
`sion and image manipulation will be discussed; the brevity of the discussions are not meant to imply that the
`technologies are not important, but that the detail required to fully understand the technologies falls beyond the
`scope of this paper.
`
`As a final and very significant part of understanding an ESC system, the impact of international standards will be
`discussed. One of the key issues is the need for a world-wide, digital, non-broadcast television-based family of
`standards for future ESC systems.
`
`In most of what follows, the emphasis will be directed toward systems that use hard copy output rather than soft
`display. The reason for this bias is based on the authors' feelings that an ESC system must produce hard copy images
`equivalent in quality to photographic prints. Many ESC images will be viewed via electronic displays, but current
`electronic display technology does not equal the photographic print or projected transparency for overall quality.
`Our crystal ball does not show us what display technology will hold sway in the future, so we, along with you, will have
`to watch the drama unfold before us.
`
`1.2 Milestones in ESC systems 3,4,5
`
`Table 1 shows a complete list ofESCs that have been developed to date. A few of them rate special recognition from
`the historical point of view.
`In 1981 Sony demonstrated its Mavica color still camera and viewer, Mavipak
`transmitter, and Mavigraph video printer. The camera had a 280,000 pixel CCD sensor with red, gTeen, and cyan
`stripes and the printer used thermal dye transfer technology with a 512-element heater. In 1986 Canon beg-an
`marketing its RC 701 ESC system in the U.S. The camera used a CCD sensor with 380,000 pixels. In 1988 Canon
`
`158 I SPIE Vol 1082 Applications of Electronic Imaging (1989)
`
`Ex. GOOG 1023
`
`

`

`introduced a high resolution version of its ESC, RC 760, with a CCD sensor that has 600,000 pixels. Also in 1988 Fuji
`Photo Ltd. demonstrated its 400,000 pixel ESC, DS-lP, that employed a removable static, random access memory,
`S-RAM card as the storage medium rather than the SVF disk. Polaroid demonstrated a monochrome ESC/motion
`camera which recorded still images on S-VHS-compact cassettes. The other ESC systems use the SVF disks for
`image storage. As can be seen from Table 1, the number of product entries are growing at a very rapid rate.
`
`Electronic Still Camera
`
`'
`
`MEDIA
`MODEL
`DATE
`Mavica
`Aug 1981 Sony Mavica
`disc
`Dec 1983 Toshiba
`July 1984 Canon
`SVF
`Oct 1984 Copal CV-1
`SVF
`SVF
`Nov 1984 Hitachi
`SVG
`Nov 1984 Panasonic
`Oct 1985 Sanyo
`SVF
`Oct 1985 Mitsubishi
`SVF
`SVF
`May 1986 Canon RC 701·
`Sep11986 Panosonic 31 oo SVF
`Nov 1986 Chinon
`SVF
`Dec 1986 Casio VS.101
`SVF
`Feb 1987. Rollei
`SVF
`May 1987 Sony MVC-A7 AF SVF
`June 1987 Konica KC 400
`SVF
`June 1987 Kodak
`SVF
`Sept 1987 Fuji ES-1
`SVF
`Nov 1987 Minona 56-90
`SVF
`Jan 1988 Konica KC 100
`SVF
`Jan 1988 Chinon
`SVF
`Mar 1988 Canon RC 760
`SVF
`Sept 1988 Nikon QV-1000C SVF
`Sept 1988 Fuji ES20
`Hi-SVF
`Sept 1988 Canon Q-PIC
`Hi-SVF
`Sept 1988 Olympus V-100 HI-SVF
`Oct 1988 Konica KC300
`HI-SVF
`Oct 1988 Matsushita ES1 0 Hi-SVF
`Oct 1988 Sony 2MVC-C1 Hi-SVF
`Oct 1988 Canon RC-470 Hi-SVF
`-Oct1988 FujiDS.1P
`RAM
`Oct 1988 Polaroid S/V-M
`S.VHS
`Nov 1988 Minolta
`Hi-SVF
`
`IMAGER
`
`LENS
`
`COMMENTS
`
`213" 280K 16-64mm
`213" 200K
`213 • 400K 16-64mm
`213 • 280K 9-27mm
`213"190K
`213" 300K 14-24mm,f/2
`213" 280K 9-27mm
`
`First ESC demo
`64mmdisc
`Demo Camera
`Demo Camera
`Demo Camera
`Demo Camera
`Demo camera
`Demo Camera
`First ESC sold in US
`
`213" 380K 11-66mm
`213" 300K 1 0&25mm, AF
`213" 250K 12-72mm,f/1. 7 Demo Camera
`213" 280K 11 mm f/2.8
`·
`Camera back for 300113
`
`213" 380K 12-72mm
`213" 300K 12-36mm,AF
`213"280K
`213" 380K x3 Zoom
`213"380K
`213" 300K 11 mm f/2.8
`
`213"600K
`213" 380K x4/x11 Zoom
`213" 400K 12-25mm, AF
`1/2"380K 11mmfl2.8
`112" 380K x3 Zoom 112.8
`112" 300K 12mm f/2.8
`1/2" 380K
`tela-wide
`112" 280K auto focus
`1/2" 380K 9& 16mm
`213"400K 16mm
`213" SSOK 12mm f/1.3
`213"380K
`
`Demo Camera
`
`Maxxum camera back
`Binocular style
`CP-9AF Camera back
`
`BtW Photojournalism
`$1400 list price
`$7001istprice
`Binocular style
`$7001istprice
`Commercial use ($2050)
`$6501istprice
`Business use (1950)
`16 MByte RAM card
`B1W still & Motion
`$1600 list price
`
`Figure 3. Functional outline of a hybrid imaging
`system and an electronic still camera system.
`
`Table 1. Electronic still camera systems that have
`been demonstrated or placed on the market.
`
`2. SYSTEM ANALYSIS
`
`The end user, the customer, of an ESC will measure the quality of the system by how well the images it produces
`compare to the images that he or she can currently obtain from a conventional 35 mm film-based system. In this
`section the foundation will be laid for how to analyze the ESC system, and the results will be used in subsequent
`sections to demonstrate the importance of the separate technologies to the final image.
`
`Figure 4 shows an image quality polygon. It is an attempt to graphically show the magnitude of the quality of each of
`the major components of the ESC system. The radial, outward spokes indicate the level of quality of each of the
`components normalized by some convenient scaling factor.
`
`A short definition of each term will now be given.
`
`1. Resolution: for an ESC this is usually defined by the number of pixels per image sensor, but film systems
`usually use the modulation transfer function, MTF, to define image resolution and sharpness, which is more
`accurate.
`
`2. Sensitivity: this is equivalent to film speed and can be expressed as an equivalent ISO speed or the minimum
`illumination (typically measured in lux) required to capture a high quality image.
`
`3. Exposure Latitude: this is the range in exposure over which the ESC can record subjectively determined
`high quality images; the exposure latitude can be expressed in terms of a ratio, for example, 400:1, in terms of
`stops, about nine, or in absolute terms, 20 lux to 8000 lux.
`
`SPIE Vol 1082 Applications of Electronic Imaging (1989} / 159
`
`Ex. GOOG 1023
`
`

`

`4. Dynamic range: this refers to the effective exposure range ofthe image on the sensor that the ESC system
`can reproduce. If the exposure control mechanism of the ESC sets the operational point in the middle of the
`possible exposure latitude, 4010 lux in the example above, the dynamic range would be plus-or-minus 3900
`lux, or a sensor dynamic range of 400:1, or a little less than nine stops. However, the best CRT displays can
`properly display little more than a 40:1 contrast ratio. This then limits the dynamic range of the ESC system
`to a little more than five stops. There is a direct parallel in the photographic system; while a color negative may
`have over ten stops in usable latitude, the system, including the limiting paper exposure latitude, may have a
`net dynamic range between four and five stops.
`
`5. Tone Reproduction: this is a measure of how well the final image gives the same appearance of the original
`scene in terms of overall contrast, shadow detail, and highlights. The physically measurable tone scale which
`produces the best subjective tone scale will vary depending on the viewing conditions; thus what is best for a
`soft display will not be the same as what is best for a reflection print, which is in turn different from what is best
`for a projected transparency.
`
`6. Color reproduction: this refers to accurate reproduction of perceived color. While it is ideal to reproduce
`exactly the perceived color of the original scene, this is neither feasible nor required. The most important
`aspects are to have good flesh-to-neutral balance, proper hue and good saturation for the basic memory
`colors, such as grass, blue sky, etc., no obvious color shifts, and no pronounced holes in the color reproduction
`space.
`
`7. Artifacts: these are unnatural occurrences in the image introduced by the various components of the ESC
`system. Two of the most obvious are due to the aliasing introduced by the low spatial sampling of the image
`resulting in too few pixels and quantization distortions which can occur if the the amplitude of the signal is
`recorded with too few bits; the resulting contours are easily seen and very displeasing.
`
`8. Noise: in solid state image sensors this is usually quantified by the non-image electrons associated with the
`sensor, output of the sensor, and its support electronics. The noise will appear as random noise or grain in the
`final print. From the point of view of a television engineer noise is measured as the ratio of the peak amplitude
`level, in volts, of the desired signal to the root-mean-square (RMS) average of the noise. This ratio is often
`expressed in decibels which is 20 times the log to the base ten of the ratio.
`
`Tone Reproduction, T
`
`Resolution, R
`
`Dynamic Range, DR
`
`MTF
`
`Color Reproduction, C
`(A)
`R
`
`c
`(B)
`
`as
`
`as
`
`04
`
`03
`
`at
`
`0.1
`
`Normalize at MTF=O.S
`Df= 1.9
`f=SOclmm
`D=0.012mm
`
`Nyquist 1
`
`Frequency
`42 clmm
`
`Spatial frequency
`
`Figure 4. The quality polygon. A. The length of the
`radial arm is proportional to the quality of the
`designated characteristic. B. The quality charac(cid:173)
`teristics are not independent and if the sensitivity
`is increased there may be a drop in resolution
`(sharpness) and an increase in visual artifacts.
`
`Figure 5. A method to calculate the effective pixel
`size that can be associated with photographic film
`when a frame transfer device with square pixels is
`assumed as the sensor model. The curve shown
`for the film is based on a theoretical model and
`does not represent a particular film.
`
`160 / SPIE. Vol. 1082 Applications of Electronic Imaging (1989)
`
`Ex. GOOG 1023
`
`

`

`The image quality polygon can be used to compare systems as shown in Figure 4. Here system A has higher
`sensitivity than system B, but the price paid for the increased sensitivity is that the resolution is lower and there are
`more artifacts. As will be demonstrated in the following sections, these image quality parameters are not independ(cid:173)
`ent. Within any given set of technologies there are trade-offs to be made in the system design.
`
`2.1. Resolution and aliasing 6, 7 ,8,9
`
`A frequently asked question about electronic sensors is how many pixels are required to produce the equivalent
`image quality of film. In this section a systematic approach is used to address this question in a system context.
`However, before the system approach is used it is instructive to look at film and the imaging sensor as individual
`components in order to understand some basic differences. Also, it will be assumed that the sensor is a CCD.
`
`Film's sharpness characteristics are usually quantified by its MTF which is obtained by exposing the film with sine
`wave targets of increasing frequency. The resulting sine wave response curve, as shown in Figure 5, is called the MTF
`and is treated as if it came from a normal linear system. In fact film is not a linear system, but no loss of
`understanding will ensue by making this approximation. Also, all color films have at least three different color layers,
`usually sensitive to blue, green, and red light. Due to increased optical scattering as the light travels deeper into the
`film, the bottom layer has a lower MTF than the top layer. For these discussions it will be assumed that a visual
`average of the three MTFs is being used; the observer views only the final print, but since the print image is a product
`ofthe dyes in the negative, the dyes in the print material, and the spectral sensitivities in the print material, all three
`must be taken into consideration in determining the visual average. An imaging site, or pixel, on a solid state sensor
`can have any shape, but here it will be assumed that it is square. The MTF of a square, ideal pixel is given by
`
`MTF sensor<D =
`
`sin( 1rDf )/( 1rDD,
`
`(1)
`
`where Dis the the pixel width and fis the spatial frequency. One obvious criteria would be to demand that the sensor
`pixel have about the same MTF as the film. 10 Figure 5 shows one such way to equate the film and pixel by
`normalizing the two curves at the 50% response frequency, f0 . This leads to the following relationship:
`D = 1.9/ ( 7Tfo).
`
`(2)
`
`Thus, if f0 = 50 cycles per millimeter, c/mm, then D = 0.012 mm. If one assumes that the the sensor is a frame
`transfer device where the pixels cover the entire surface of the sensor and that the sensor has the same aspect ratio as
`the 35 mm frame, 2:3, then the sensor must have 2000 by 3000 pixels. To date, sensors with six million pixels have
`not been reported.
`
`The above analysis is not complete, however. The CCD is a sampling image sensor while film is a continuous sensor.
`The nature of sampling introduces aliasing. Aliasing is the generation of false signals or, in the case of image sensors,
`false images. Figures 6, 7, and 8 demonstrate this property of sampled systems. If the sine waves in Figure 6 are
`sampled at a rate less than twice the frequency ofthe sine wave, the resulting signal will appear as a lower frequency
`or aliased signal. Figure 7 shows the spectra of the four sine waves. Note that the fourth and seventh harmonics are
`aliased to lower frequencies. Figure 8 shows the case of a simple image spectrum. When the sampling is high enough,
`above the Nyquist frequency which is equal to twice the highest frequency one wishes to record faithfully, no aliasing
`takes place but, as shown in part b of Figure 8, when the sampling frequency is below the Nyquist frequency, the
`spectra will overlap, giving rise to many artifacts. Figure 9 shows what happens when an image with strong vertical
`lines is sampled with an imager that does not meet the Nyquist sampling criteria; note the strong low frequency
`banding. This problem becomes acute in sensors that have color filter arrays on them, for then the banding can
`become a rainbow due to the different relative phases of the colored pixels on the sensor.
`
`Thus in designing a system that samples an image, one must try to avoid or minimize the problems introduced by
`aliasing. In theory, the simplest method is to sample at twice the frequency of the zero, or near zero, response of the
`film. In the example shown in Figure 5 this means that we should sample the image, relative to the image plane, at
`about 300 times per millimeter since the MTF at 150 c/mm is 10%. This would require a format of7,200 by 10,800
`pixels or 77.6 million pixels. Clearly, while this solves the aliasing problem, it is an unacceptable, if not impossible,
`solution. As will be discussed in the next section, aliasing can be reduced but at the price of sharpness.
`
`SPIE Vol. 1082 Applications of Electronic Imaging (1989) / 161
`
`Ex. GOOG 1023
`
`

`

`/ \ / \ 1 /41
`
`1121
`
`"
`
`v~ "
`f\Af\1\
`vvv v
`nnAAAAAHAH
`vvvrmrvm "'"•
`AMWWAWA
`vmvvvvvvmv '14
`
`Amplitude
`
`Nyquist Frequency
`
`Nyquist Frequency
`
`Resolved
`
`Resolved
`
`Aliased
`
`Aliased
`~To~
`To~~
`' ' '
`
`i 4
`
`Frequency
`t
`'·- ux
`
`'"
`
`(b)
`
`Figure 6. Aliasing. A. Four sine wave curves are sam(cid:173)
`pled at very high resolution and no aliasing takes
`place. B. The same four curves are sampled at low
`resolution and the sine waves with the two highest
`frequencies are aliased to lower frequencies.
`
`Figure 7. The Fourier transform of the curves in
`Figure 6. Note that the aliased, high frequency
`signals are reflected, relative to the Nyquist fre(cid:173)
`quency, to lower frequencies.
`
`, , ,
`
`v , '
`, '
`'
`'
`,
`,
`f = -ik -fN Q
`
`(b)
`
`, '
`,
`'
`
`y , '
`,
`'
`
`' ' '
`
`Figure 8. The impact of aliasing on image spectra. A.
`In this case the sampling distance, ~X. is small
`enough to ensure that the entire spectrum will lie
`within the Nyquist frequency and be faithfully re(cid:173)
`corded. B. In this case the sampling distance is large
`and the image spectrum lies beyond the Nyquist
`frequency, resulting in a strongly aliased signal.
`
`162 / SPIE Vol 1082 Applications of Electronic Imaging (1989)
`
`Figure 9. A computer simulated sample of aliasing
`from a stripped imaging sensor with a resolution
`of 500 by 720 pixels.
`
`Ex. GOOG 1023
`
`

`

`2.2
`
`System sharpness and CMT acutance 11
`
`The above analysis does not properly consider how the observer "sees" the total photographic system. The customer
`looks at the final print, not at the CCD sensor, and it is there where one should analyze what is required. After the
`analysis has been made on the final print (system) MTF, one can check on the implications of aliasing.
`
`The quality target for ESC is the film system, and it is composed of a camera lens, the film, the printer lens and the
`paper. Each of these components can be described, at least to a good approximation, as a linear system and thus by an
`MTF curve. The same can be said for an ESC system, except here the sensor replaces the film and some sort of
`electronic printing device replaces the printer lens and paper of the film system. For this analysis the ESC system
`printer will be modeled as a laser beam writing onto photographic paper; for reasons that will be discussed in the
`section on hard copy it is assumed here that the printing process uses twice the number of lines used in the image
`sensor. In the analysis that follows it is assumed that the camera lens and paper are the same in the two systems. Also
`it will be assumed that the sensor is the same size as the film, although in practice this would not be the case, and that
`all system evaluations will be made in the paper plane. If it is assumed that the final print is 4 inches by 6 inches,
`10.16 em by 15.24 em, then the system magnification is about 4.23. Thus the film, sensor, and lens MTF's must be
`adjusted by this 4.23 magnification factor to properly carry out the evaluation process in the paper plane.
`
`Figure 10 shows the individual component MTF's of a film system along with the cascaded system MTF. Figure 11
`shows the same for an ESC system with a CCD sensor with a format of 2000 pixels by 3000 pixels. The two system
`curves are different enough to make it difficult to assume that they would have equivalent quality. One of the
`standard methods in evaluating photographic images is using CMT acutance as developed by Crane 12 and
`Gendron.13 Here a slightly altered form of CMT acutance will be used and the emphasis will be on relative quality
`rather than absolute quality. In short, if two systems have the same CMT acutance, it can be assumed that they will
`look equally sharp, but they may not have equivalent resolution. When either a film or an ESC system uses strong
`nonlinear image enhancement, the value of CMT acutance becomes questionable, but for the discussions here this
`will not be a problem. 10 For a rough calibration, images that have a CMT value above 80 are acceptable for image
`sharpness, and images with CMT values above 92 are excellent in terms of sharpness.
`
`FILM
`
`CMT=89.1
`
`06
`MTF
`
`Q2
`
`ESC
`
`CMT=91.6
`
`0.8
`
`0.6
`MTF
`04
`
`0.2
`
`0o~~2_L~4~-e~~-L~,o~=,~2~~,4---1~6~~~
`Frequency in paper plane
`
`Figure 11. The MTF curves of an electronic still
`Figure 10. MTF curves of a film system consisting
`camera system consisting of a camera, sensor,
`of a camera, film, printer and paper. The CMT
`laser beam printer, and paper. The sensor has
`value is obtained by cascading the eye MTF curve
`2000 by 3000 pixels.
`with the system MTF curve.
`The system MTF, MTF s (0, is the cascaded product of the individual component MTF's, MTF i (0;
`
`SPIE Vol 1082 Applications of Electronic Imaging (1989) / 163
`
`Ex. GOOG 1023
`
`

`

`(3)
`
`where the components of the system include the camera lens, the film or sensor, the printer lens or laser beam, the
`paper, and the eye.
`
`If MTF eye (f) is the MTF of the eye at a viewing distance of four picture heights from the print, about 40 em for the
`4-inch print, then the CMT acutance is defined by
`CMT = 100 + 66LOG( R ) ,
`
`(4)
`
`where R is given by,
`R = fXJ MTFeye(f) MTFs(f) df I Loo MTFeye(f) df .
`
`(5)
`
`In the above examples the film system CMT acutance is 89.1 while that of the ESC system is 91.6. The reason for the
`ESC having a higher CMT acutance value is due to the better low frequency response of the square pixel and the
`doubling the printing line rate as previously mentioned.
`
`If there are 3000 pixels in each line, then the Nyquist limit is about 10 c/mm, which is beyond what the typical
`observer can see at 40 em. However, the pixel MTF is such that up to 20 c/mm is recorded and thus high frequency
`information can be aliased below the 10 c/mm limit; for example, a frequency of 16 c/mm in the paper plane will be
`aliased to 4 c/mm in the print and can easily be seen, while a strong signal near 19 c/mm in the paper plane will be
`aliased down to 1 c/mm on the print, which would be seen as a serious artifact. Thus it is important to optically
`pre-filter the sensor image in such a manner that high frequency information is not aliased by the sensor into the
`visible spatial frequency region. For the given system magnification this means that information beyond the Nyquist
`frequency of 42 c/mm in the plane of the sensor should be heavily attenuated; see Figure 5.
`
`Figure 12 shows the situation when a lens has been chosen that reduces much of the signal beyond 42 c/mm in the
`plane of the sensor or 10 c/mm in the plane of the paper. The CMT acutance value is now 88.8, a little less than that of
`the film system. Thus with sufficient optical prefiltering it is possible to produce a very sharp image with a
`2000-by-3000 pixel sensor with no aliasing. Figures 13 and 14 show the family of system MTF curves as a function of
`
`ESC= PRE-FILTER
`
`Nyquist
`Frequency
`
`CMT=88.5
`
`NO OPTICAL PRE FILTERING
`
`0.6
`MTF
`
`04
`
`U2
`
`System
`MTF OS
`
`0.1
`
`02
`
`~~~-2~~~4~~~6~~~8~~~~~12
`Spatial frequency (c/mm) .
`
`Figure 12. The same ESC system as shown in
`Figure 11 but with optical pre-filtering to reduce
`aliasing.
`
`Figure 13. A series of ESC system MTFs of vary(cid:173)
`ing resolutions with no optical pre-filtering.
`
`164 / SPIE Vol. 1082 Applications of Electronic Imaging (1989)
`
`Ex. GOOG 1023
`
`

`

`the number of pixels per picture height. In Figure 13 no anti-aliasing optical pre-filtering is introduced, and it is seen
`that the ideal1500- and 1000-pixel sensors can produce 4-inch by 6-inch print MTFs that are close to that of film,
`but the 750-pixel and 500-pixel sensors will produce substantially poorer images; an observer can see a one CMT
`acutance unit change. However, the 1000-pixel sensor will have strong aliasing; any information above 7.5 c/mm will
`be aliaseq, so if the original scene has any information around 14 c/mm in the paper plane or 60 c/mm in the sensor
`plane, a strong aliased signal of 1 c/mm will appear, and this is at the peak of the eye's spatial frequency for the
`defined viewing conditions.
`
`Figure 14 shows the same set of curves but when the proper optical pre-filtering is applied to prevent the aliasing. As
`can be seen from Figure 14, the quality ofthe image drops off rapidly as the pixels per sensor height moves from 2000
`to 500. Hence, while a 2000-line sensor is capable of producing print MTFs as good as or better than the defined film
`system, a 1500- (and below) line system will require some digital signal processing to produce a print MTF equal to
`the film system. The effects of digital image processing will be covered in a later section. The optical pre-filtering
`used in these calculations are not optimal in that they have a gradual cut-off characteristic which significantly
`attenuates low spatial frequency information. By using more sophisticated optical methods, like birefringent blur
`filters, a sharper cut-off characteristic can be obtained which allows more of the lower spatial frequency information
`to be recorded at higher levels of modulation. This results in -better quality prints.
`
`OPTICAL PRE FILTERING
`
`System
`MTF
`
`06
`
`011
`
`1.2
`
`c .g 0.6
`.,
`:g 011
`0
`:::; 0.2
`
`c· 8x10 MTF-F, film
`• · 3R MTF-F, film
`•-SVF MTF
`o- 3R MTF, mega pixel sensor
`• • 8x10 MTF, mega pixel sensor
`
`Figure 14. The same series as in Figure 13, but
`with optical pre-filtering to reduce the effects of
`aliasing. The sharp drop in CMT values can be
`reduced by using optical pre-filters that have
`sharper cutoff characteristics.
`
`Figure 15. Experimental results. The SVF system
`has a very poor system MTF relative to the photo(cid:173)
`graphic system or the system designed around a
`Kodak sensor using 1.4 million pixels. The elec(cid:173)
`tronic enhancement greatly improves the high
`resolution ESC system prints.
`
`There is often a desire to make enlargements, commonly up to 8 inches by 12 inches and sometimes much larger,
`from a 35 mm negative. In general, if one views the enlargements from a greater distance so that the viewing angle is
`the same as when viewing the smaller prints, the images will appear to be sharper; the reason for this is that the paper
`is able to more faithfully record the image since the high spatial frequency information in the film or sensor has been
`shifted to lower spatial frequencies on the print, that is, one uses more paper for the same image. However, often an
`observer will look at an enlargement from about the same distance that he or she looks at a small print. The effect is
`to see more of the fine detail, and since it is reproduced at a lower level of modulation than the lower frequency
`information, the image looks poorer. For example, if the film system is enlarged 8.5 times to produce a 8-inch by
`12-inch print and it is viewed from about 40 em, its CMT acutance is 81.3. The 2000-pixel-per-sensor height image
`has a CMT acutance of 84.7 without optic