An improved logarithmic response pixel offering high speed and linear response at low light levels, as well as increased voltage swing in the logarithmic regime is presented.

The availability of optical MEMS (Micro-Electro-Mechanical Systems) is promising to revolutionize optical instrument design. We are developing a multiobject imaging spectrometer based on a commercially available optical MEMS--the Texas Instrument's Digital Micromirror Device (DMD)^TM. The micromirror array is laced at an image plane of an optical system, and is used as a spatial light modulator to redirect portions of the image into the spectrograph. The programmability of the micromirror array allows the creation of arbitrary `slit' patterns as input to the spectrograph. In addition, by controlling the dwell time of each micromirror individually, it is possible to adaptively extend the dynamic range of the spectral imaging system.

Sheet-of-light range imaging is an efficient technique for measuring range. A sheet of light is projected onto an object, and the cross-sectional line of the light sheet and the object is recorded with an image sensor. Range is computed by means of triangulation. In this paper we discuss the design and test results on an image sensor specifically designed for sheet-of-light. The sensor consists of an array of rectangular Position Sensitive Detectors (PSD's), combined with an on-chip multiplexer (a `smart' PSD array). A PSD delivers two photocurrents, the ratio of which is a measure for the center of gravity of the light intensity distribution over the PSD. Measurements have shown that the smart PSD-array delivers 2 million range values per second, at a 1:500 resolution. The smart PSD-array consists of an array of PSD's with a 28 micron pitch, and a length of 20 mm. A DSP board takes care of computation of the range data from the PSD signals.

Conventional CCD image sensors have low responsivity in the ultraviolet region of the spectrum because the small absorption depth results in electron generation in the gate of the CCD rather than in the buried channel. Inorganic phosphor coatings to convert UV to visible light have been developed through the collaborative efforts of the University of Waterloo and DALSA Inc. In general, the coatings are simple to deposit and have been used to obtain images at illumination wavelengths of 265 and 365 nm. The developed coating is similar to acrylics doped with organic laser dyes presented in previous papers; here, the organic dye has been replaced with an inorganic phosphor material. Organic laser dyes have the advantage of nanosecond to microsecond decay times and the availability of a variety of dyes enables the customization of the absorption and emission wavelengths, however, they quickly degrade upon exposure to UV radiation. In many manufacturing applications, higher UV illumination intensities will degrade the laser dyes in a relatively short time. Inorganic phosphors exhibit reduced degradation rates thereby improving the viability of phosphor coated CCD image sensors for commercial and manufacturing applications. The quantum efficiency improvement of the coated sensor over the uncoated sensor is 350% and 86% at 265 and 365 nm, respectively.

DALSA and Philips Digital Video Systems have developed a high performance pinned-photo-diode CCD linescan sensor for scanning of motion picture film. This application requires a sensor with performance optimized for a high dynamic range optical input, with high speed operation. The sensor is named the IL-P2-2048. It is based on the existing DALSA IL- P1-2048 linescan sensor design with a number of significant changes--the full well has been increased to in excess of 300,000 electrons, with a maximum video swing of more than 1000 mV with two bi-directional outputs operating at > 21 MHz video rate. Of importance is the introduction of pinned- photo diode pixels (10 X 10 micron pitch) which possess extremely low fixed pattern noise, low pixel-response non- uniformity, and low image lag. The sensor also has pixel select gates to enable the user to select the full resolution or a reduced number of pixels for use in different imaging formats; the pixel selection is nevertheless dynamic and can be user selectable. The output circuitry has excellent linearity from a few mV to the full signal swing of 1000 mV. In this paper the authors present the design and test results of the sensor.

We have developed a linescan sensor suited for high image quality, high-resolution, high-speed imaging. The 6k-pixel sensor has: four corner outputs each operating at 40 MHz for high scanning speed; 7 micrometers pixels for shorter sensor length and simpler optical design; an exposure control and antiblooming structure that does not produce imaging artifacts; 9.5 (mu) V/e charge conversion efficiency at the output for enhanced sensitivity and dynamic range under light-starved conditions; optimized pinned photodiodes for low image lag (< 350 electrons) and enhanced UV response (> 40% QE at 250 nm); 100% fill factor down to the deep UV; a pixel storage structure that suppresses photosite-to- shift-register optical crosstalk; highly linear output structures and amplifiers (< 1% non-linearity); matched 5- V 2-phase clocks that can be driven with off-the-shelf CMOS drivers; output waveform shape that allows 40-MHz CDS; and photoresponse non-uniformity that is < +/- 2% of the signal.

The CFH12K is a 12k X 8k CCD mosaic camera for the Canada-France-Hawaii Telescope (CFHT), a 3.6 m telescope located on Mauna Kea, Hawaii. The CFH12K is comprised of twelve 4k X 2k thinned backside-illuminated CCDs, arranged in a close-packed array of two rows each containing six CCDs. Located at the CFHT Prime Focus (f/4.2), the CFH12K provides a 42 by 28 arcminute field-of-view, 0.206 arcsecond per pixel sampling, with a resulting data file of more than 200 Mbytes per image. The camera has been designed to exploit the exceptional wide-field imaging capability provided by the CFHT. At the time of it's commissioning in January 1999, the CFH12K is the largest thinned close-packed CCD mosaic in astronomy. This paper describes the system architecture, and some of the relevant issues associated with the construction, evaluation, and operation of very large mosaic cameras. Emphasis is given to system design issues, illustrating the CFH12k as part of a larger system: the CFHT.

We have developed two single-chip CCD sensor architectures for high-speed, 3-channel color imaging. Both are line-scan sensors for Time Delay and Integration (TDI) imaging. One architecture achieves a sub-microsecond TDI register shift time by contacting metal to poly-Si gates through the imaging regions. The other has no metal in the imaging regions and requires a longer shift time. Both sensors are capable of 40 MHz data rate per channel. Line rates for 2048-pixel devices of 16.5 and 18.5 kHz (shift times of 7.5 and 0.7 microsecond(s) /stage) are achieved.

In July of 1998 the National Optical Astronomy Observatories (NOAO) successfully upgraded MOSAIC 1, an 8192 by 8192 pixel array using eight Scientific Imaging Technologies, Inc. (SITe) St-002A thinned backside 2k by 4k charge coupled devices (CCDs). In July of 1999 MOSAIC II, a clone of MOSAIC I was commissioned also using eight SITe ST-002A CCDs. Additionally in December of 1998 NOAO implemented Mini- MOSAIC a 4096 by 4096 pixel array using two SITe ST-002A thinned CCDs. This report will discuss the performance, characterization and capabilities of the three wide field imagers now in operation at NOAO's Kitt Peak Observatory, Cerro Tololo Inter-American Observatory and at the WIYN Consortium 3.5-Meter telescope on Kitt Peak.

The first section will review the requirements for scientific/industrial cameras and discuss the limitations of conventional CCD and active pixel sensors (APS) approaches. The next section will describe the Active Column Sensor technology, and discuss the reasons for the improved performance compared to passive pixel CMOS and APS imagers. Then the Active Column Sensor^TM (ACS) pixel structure and modes of pixel operation will be discussed. The paper will also describe the other camera functions that are placed on the same substrate. Experimental results including images will be presented. The paper will close with a glimpse into the future of industrial and scientific CMOS ACS image enabled systems.

This paper describes a 128 X 128 pixels prototype array organized as sub-arrays of 32 X 32 pixels each, with 21 micrometers pixel pitch. The sub-arrays, photodiode or photogate based, are implemented using a standard 0.7 micrometers CMOS process. Various topologies of the photosensitive area have been implemented and some of them have an optical metal shield over the so-called non-sensitive area to evaluate the contribution of the active electronic area to the responsivity of the sensor. A synthesis of the measurements carried out by CIMI-SUPAERO and MMS, addressing darkness parameters, noise, photometric and radiometric performances, are presented with emphasis on the photogate type pixels. Results of spot-scan analysis and crosstalk measurements performed on selected topologies are also reported. Several samples were irradiated at different proton doses and their related behavior is discussed. From these results, a new 512 X 512 pixels array has been designed for space applications. The main features of this APS device are presented here.

Pixel reset noise sets the fundamental detection limit on photodiode based CMOS image sensors. Reset noise in standard active pixel sensor (APS) is well understood and is of order kT/C. In this paper we present a new technique for resetting photodiodes, called active reset, which reduces reset noise without adding lag. Active reset can be applied to standard APS. Active reset uses bandlimiting and capacitive feedback to reduce reset noise. This paper discusses the operation of an active reset pixel, and presents an analysis of lag and noise. Measured results from a 6 transistor per pixel 0.35 micrometers CMOS implementation are presented. Measured results show that reset noise can be reduced to less than kT/18C using active reset. We find that theory simulation and measured results all match closely.

In this paper a programmable imager with averaging capabilities will be described which is intended for averaging of different groups or sets of pixels formed by n X n kernels, n X m kernels or any group of randomly- selected pixels across the array. This imager is a 64 X 64 array which uses passive pixels with electronic shutter and anti-blooming structure that can be randomly accessed. The read-out stage includes a sole charge amplifier with programmable gain, a sample-and-hold structure and an analog buffer. This read-out structure is different from other existing imagers with variable resolution since it uses a sole charge amplifier, whereas the conventional structure employs an opamp per column plus another global opamp. this architecture allows a reduction of the fixed-pattern noise observed in standard imagers. The prototype also includes an analog to digital converter which provides the digital output of the images.

Co⁶⁰ irradiations have been carried out on test structures for the development of CMOS Active Pixel Sensors that can be used in a radiation environment. The basic mechanisms that may cause failure are presented. Ionization induced damage effects such as field leakage currents and dark current increase are discussed in detail. Two different approaches to overcome these problems are considered and their advantages and disadvantages are compared. Total dose results are presented on a pixel that can tolerate more than 200 kGy(Si) (20 Mrad(Si)) from a Co⁶⁰ source.

As CMOS technology scales, the effect of 1/f noise on low frequency analog circuits such as CMOS image sensors becomes more pronounced, and therefore must be more accurately estimated. analysis of 1/f noise is typically performed in the frequency domain even though the process is nonstationary. To find out if the frequency domain analysis produces acceptable results, the paper introduces a time domain method based on a nonstationary extension of a recently developed, and generally agreed upon physical model for 1/f noise in MOS transistors. The time domain method is used to analyze the effect of 1/f noise due to pixel level transistors in a CMOS APS. The results show that the frequency domain results can be quite inaccurate especially in estimating the 1/f noise effect of the reset transistor. It is also shown that CDS does not in general reduce the effect of the 1/f noise.

A new generation of smart pixels, so-called demodulation or lock-in pixels is introduced in this paper. These devices are capable of measuring phase, amplitude and offset of modulated light up to some tens of MHz, making them ideally suited to be used as receivers in 3D time-of-flight (TOF) distance measurement systems. Different architectures of such devices are presented and their specific advantages and disadvantages are discussed. Furthermore, a simple model is introduced giving the shot noise limited range resolution of a range camera working with these demodulation pixels. Finally, a complete TOF range camera based on an array of one of the new lock-in pixels will be described. This TOF- camera uses only standard components and does not need any mechanically scanning parts. With this camera non- cooperative targets can be measured with a few centimeters resolution over a distance of up to 20 meters.

Its official name is IEEE 1394. It is also called FireWire^TM (by Apple) and i.LINK^TM (by Sony who has licensed this name to several other companies). It was created in the late 1980s as a digital technology designed to satisfy the ever-increasing demand for communications bandwidth for video. It was also designed to simplify and reduce the number of cables and connectors between devices. Simple one-cable connectivity has become a necessity brought about by miniaturization and increasing device functionality as ports have reduced in size and number. The IEEE 1394 high performance serial bus has already found its way into consumer products and one day probably will link every digital appliance in the house: TV, DVR, camcorder, hi-fi stereo, computers and peripherals. Now it's poised to make the same impact in the industrial/scientific imaging world.

An ultra-high definition experimental camera system has been designed with double the horizontal and vertical resolution of HDTV. An 8M-pixel CCD with a progressive 60 frame-per- second scan-rate has been developed for the system. The 34 mm X 17.2 mm image area has 4046 (H) X 2048 (V) active imaging pixels with 8.4-micrometers squares. This CCD has a split- frame transfer structure and sixteen 37.125 MHz outputs so that the vertical and horizontal transfer frequencies are almost the same as those of HDTV. The split-frame transfer structure halves the required VCCD clock speeds and thus improves charge transfer efficiency. The multiple-output structure with its 16 outputs enables high data-rate imaging for ultra-high resolution moving pictures. In the signal processing section, analog gain adjustment circuits correct for the mismatches in the characteristics of outputs, and a correlated double-sampling technology is employed on each of the 16 CCD output signals. The output signals are digitized by 12-bit ADCs. The converted signals are then sent to the digital signal processing (DSP) circuits. In the DSP circuits, the upper half of the captured image is vertically inverted. All of the output data is then merged into a 4K X 2K pixel image and reformatted to create twenty-four 640 (H) X 480 (V) pixel sub-images for image processing. After contour compensation processing, the video signals are converted into an analog signal and presented on two ultra high resolution video monitors.

Three-chip common-optical axis `beamsplitter' color cameras may seem complex, but they can provide image acquisition that is simultaneous in time and space, is perfectly registered, and covers 100% of three color bands--in other words, they deliver the `Holy Grail' of color imaging. Alternative schemes cannot match this feat, not can they match a good beamsplitter's image fidelity or ease of system integration. A common optical axis camera with integrated processing capabilities for color balancing offers even more advantages over the alternatives, as better color registration and uniformity result in reduced system, development and maintenance costs. The prisms associated with the beam splitter add a large path length of glass between a lens and sensor, and most systems are not designed to account for this altered path. To use a single lens the system should be optimized to compensate for the chromatic aberrations introduced by the various optical elements.

We present various methods to increase the spatial resolution of an undersampled, and thus aliased, image sequence. The sequence is acquired by an infrared camera, which severely undersamples the image of a static scene. Vibration of the setup causes a random, sub-pixel, global translation of the scene before sampling. Although a single frame is hampered by aliasing, a sufficiently large series of such frames contains all the information needed to undo this aliasing.

In electronic amplifiers of image signal in television cameras is sometimes used complex antinoise correction. It effectively suppresses self-noise of the amplifier. The essence of the correction consists in including an inductive choke in the input circuit of the amplifier, due to that high-frequency impedance is increased and, together with it, the level of the signal components in this region of frequencies relative to the self-noise of the amplifier. Careful analysis of the circuits of the complex antinoise correction with the account of the losses in the input circuit has allowed to get new conclusions, which can be useful in constructing the low-noise video amplifiers.

Electron Bombardment CCD sensor (EB-CCD) had originally developed for the low light imaging application and recently it has been coming to be used as an alternative to SIT tube camera and I-CCD camera. However, some of applications required higher sensitivity and higher dynamic range. In order to meet these expectations, we have been continuing the improvement of sensor such as higher gain and higher quantum efficiency. In this paper, we would like to introduce the new high gain sensor and new photo-cathode of GaAs and GaAsP, which realize the 30 to 50% quantum efficiency. And also we would like to show that how GaAs EB- CCD is sensitive based on the sensitivity comparison test between multi alkali photo-cathode image intensifier and new GaAs photo-cathode EB-CCD sensor.

The Visually Guided Vehicle uses the newly available microvideo systems combined with a miniature radio control system to allow precision delivery of non-lethal pyrotechnic combined with a miniature radio control system to allow precision delivery of non-lethal pyrotechnic effects at ranges of 1000 to 2500 meters. The man portable device contains a rocket powered, tube launched vehicle which powers to apogee and glides downward into the desired location. This semi-powered design saves significant weight. This paper will highlight the test data, photographs, and videotape of both the electronic components used in the design and the final non-lethal effects made possible by the complete system.

The principle and method of measurement and signal processing of the vibration for the structure of interferometric optical fiber sensor are described and vibration response property of the structure is measured under the free vibration with attenuation and vibration subjected to force in this paper. The measurement results by the optical fiber sensor and electric strain gauge are proximate, so this research provides a kinds of new method for real-time vibration monitoring of engineering structure.

In the two-dimension or three-dimension object recognition research, the shapes, the sizes, and the geometry of the objects are usually compared to recognize different objects. However, the quality of the lenses of the CCD cameras is different between manufacturing company. If the quality is poor, the mapping ranges in the camera lenses itself in the X and Y directions are not equal. The mapping range discrepancy will cause the distortion of the object image after the CCD camera takes the picture of the object. In order to get the correct geometry comparison of two images; the distorted images need to be restored to its original images. This paper invests the method to correct the distorted image to its original image. In this paper, the second moment is used to find the center point and the orientation of the object. the distortion correction coefficients are introduced to the second moment equations to find the actual center point and the orientation of the object. Image rotation and correction are introduced in this paper, too. The distortion correction coefficient work well to correct the distorted image. By using the correction coefficient, the image can be restored to its original image and the image rotation equations can rotate the image to its proper position.

This paper describes an approach to implement the motion detection for a low-cost surveillance camera system. The motion detection is usually implemented by the image segmentation method in industrial applications. However, the image segmentation method needs the additional memory for segmenting and saving an image and longer processing time, which is not suitable for low-cost surveillance camera system in industrial applications. Therefore, this paper suggests the motion detection only utilizing the digital camera processor (DCP). Because the DCP is an essential part of the surveillance camera system, this method does not need the additional cost of memory.

Numerical analysis has been carried out on the spectral dependence of the quantum yield for an AlGaN(n)-GaN(p) photodiode ultraviolet (UV) detector in which the AlGaN layer is an energy band-gap graded (EBGG). The latter has been achieved by variation of Al content. The spatial dependence of the material properties, such as energy band- gap and absorption coefficient of the photodiode's n-type layer is considered in the calculation. The band-gap grading, due to a reduced absorption coefficient at the surface region and the built in electric field, results in the increase of the minority carrier generation in the vicinity of the junction resulting the enhancement of carrier collection efficiency. In order to improve the quantum yield, it was found that there is no need to fabricate the structures exhibiting the large gradients. Within the range of small values of EBGG there is a substantial rise in the detector quantum yield with increasing the grading. Much more extended the dependence in quantum yield with the grating has been found when the recombination was present at the illuminated surface. The carried out study enabled to determine the optimal grading and the graded layer thickness with regard to the detector response to the selected UV wavelengths.

In the paper, Laser Measurement System (LMS) is used to monitor region in ship lock of the Three Gorges Project. Several methods are carried out to detect objects in the region. First, in order to determine whether there are ships in detected region, a distance difference integral method is given. Second, in order to decrease the random noise affection, the neutralization and the relativity of the fore-and-aft data are used to eliminate the noise and to eliminate misjudge from the random data. Third, two LMSs are installed to trace the motive object, a fixed LMS detects a linear region, and the other LMS can be rotated to scan a plane region in ship lock. As the fix LMS detected some motive objects, the other LMS begins to rotate and trace changed region and to determine whether there are objects in detected region. Finally, Laser Measurement System is tested in SanBo ship lock in ZheJiang province of China. All the methods are very effective. And the system will be installed in permanent ship lock of the Three Gorges Project in 2003. The system will keep the safety of the lock.

This paper describes how the operating principles and assemblies associated with high-end imaging devices have been reduced to readily comprehensible concepts as well as affordable, accessible and manipulatable hardware. An inexpensive had-operated scanner was modified to perform imaging applications that are generally associated with high end equipment.

A highly accurate multispectral camera and the application software have been developed as a practical system to capture digital images of the artworks stored in galleries and museums. Instead of recording color data in the conventional three RGB primary colors, the newly developed camera and the software carry out a pixel-wise estimation of spectral reflectance, the color data specific to the object, to enable the practical multispectral imaging. In order to realize the accurate multispectral imaging, the dynamic range of the camera is set to 14 bits or over and the output bits to 14 bits so as to allow capturing even when the difference in light quantity between the each channel is large. Further, a small-size rotary color filter was simultaneously developed to keep the camera to a practical size. We have developed software capable of selecting the optimum combination of color filters available in the market. Using this software, n types of color filter can be selected from m types of color filter giving a minimum Euclidean distance or minimum color difference in CIELAB color space between actual and estimated spectral reflectance as to 147 types of oil paint samples.

A digital preprocessor for the real-time compression of raw image data is currently under development. Since high input data rates (<EQ 160 Mb/s) have been specified, the processor is being developed fully in hardware. This hardware realization (ALTERA^TM CPLD and space-qualified ASIC technology) implements a compression scheme based on a Discrete Wavelet Transform and a Successive Approximation Quantization coding scheme. Therefore compression can be applied either in fixed rate or in fixed error modes, including a lossless mode.

This paper describes a hybrid technique of coupling a through-the-lens band-pass illumination sensor with a software segmentation and histogram algorithm to estimate the scene illuminant(s) in a digital still camera and color correct the resulting image. A fuzzy logic algorithm combines the result of the source illumination(s) with the chromaticity determination of near-neutral objects derived from the image pixel data to determine the amount chromatic correction required.

In consumer digital cameras, some of the primary tasks in the image capture data path include automatic focus, automatic exposure determination and auto-white balance (AWB). There are numerous algorithms used in implementing these tasks--auto-focus is implemented using maximum contrast, ranging or sonar; white balance using color gamut determinations and `gray value estimations', and auto- exposure using scene evaluations. We evaluate the system implications of implementing one of these tasks, namely white balance on an embedded system--a digital camera. There include, among other things, design approach, power consumption, software vs. hardware implementation and microprocessor vs. ASIC implementation. Commercially available digital cameras and their choice of AWB implementation are discussed where appropriate. Such an evaluation will assist, we hope, anyone designing or building a digital camera sub-system.

In the following, an analytical model for the optical imaging will be presented that completely describes the optical transfer pathway from the original object to the in- focus as well as to the out-of-focus projection. According to this, both the diffraction theory and the Gaussian beam theory (geometrical optics) will be taken into account. The results show that noticeable differences between these theories are found near the focal plane only. The model is generally applicable and therefore valid for projection type displays and acquisition devices (cameras).

Modeling color imaging sensors requires the computation of the amount of electric potential collected within each well of the sensor. This amount is proportional to the integral of the product of the quantum efficiencies of the sensor and the spectral representation of the color of the incident light. Although the quantum efficiencies of any color imaging sensor are available from the data sheets as a function of wavelength, the spectrum of light is generally expressed through a small number of its components along the color matching functions. Since an exact reconstruction of the spectrum of light is not possible from these components, various approximations are used for its reconstruction. In the simplest case, the color matching functions are taken as constants over nonoverlapping regions of the spectrum and as zero outside these regions. The widely used color matching functions are used in the second approach where the components of the spectrum of light along these functions are available. There components are computed in the third approach taking the nonorthogonality of the color matching functions into account. The effects of these approximations are demonstrated with results from the simulation model of a color imaging sensor.

In this paper, we review several algorithms that have been proposed to transform a high dynamic range image into a reduced dynamic range image that matches the general appearance of the original. We organize these algorithms into two categories: tone reproduction curves (TRCs) and tone reproduction operators (TROs). TRCs operate pointwise on the image data, making the algorithms simple and efficient. TROs use the spatial structure of the image data and attempt to preserve local image contrast.

We present a new digital image interpolation technique and analyze its use in digital photography. Our main goal is generate interpolated images with high quality, eliminating or drastically reducing distortions caused by reconstruction error. Our initial hypothesis is that interpolation techniques with frequency response closer to ideal reconstruction can reduce or eliminate reconstruction error, improving image quality. The proposed technique is a new 2D third-order filter, implementing discrete convolution between image samples and interpolation coefficients that are different for each interpolated pixel. To validate the proposed technique, we compared it with piecewise constant, linear and cubic interpolators.

integrated Circuits (IC) and other sophisticated electronics consumer goods are commonplace in cameras of all kinds, sizes and price range. Today's state of the art digital imaging now makes it possible to achieve photographic quality at levels sufficient to support the growth of cameras applications such as video conferencing, video phones and Video Cellular phones, video Email and WEB site imaging. Additionally, to generate mass market appeal, digital imaging technology integrators and camera manufacturers had lowered the price of their products, increasing the pressure to lower the cost of the imaging engine, including both the die and the package. A newly introduced Die Size, Opto-Electronic Package offers a substantial reduction of both package costs and size (1,2). It provides the smallest and thinnest optical package for solid state light detectors and imaging devices such as CCD's and CMOS, and enables new applications which could not be realized with the current packaging technology. Colored imaging devices, in a variety of sizes have been packaged (figure 1). Optronic CSP's offer excellent dimensional accuracies with very tight mechanical tolerances.

CMOS image sensor designers take advantage of technology scaling either by reducing pixel size or by adding more transistors to the pixel. In both cases, the distance from the chip surface to the photodiode increases relative to the photodiode planar dimensions. As a result, light must ravel through an increasingly deeper and narrower `tunnel' before it reaches the photodiode. This is especially problematic for light incident at oblique angles; the narrow tunnel walls cast a shadow on the photodiode, which in turn severely reduces its effective QE. We refer to this phenomenon as pixel vignetting. The paper presents experimental results from a 640 X 512 CMOS image sensor fabricated using a 0.35(mu) 4-layer metal CMOs process that shows significant QE reduction of up to 50% for off-axis relative to on-axis pixels. Using simple geometric models of the sensor and the imaging optics, we compare the QE for on and off-axis pixels. We find that our analysis results support the hypothesis that the experimentally observed QE reduction is indeed due to pixel vignetting. We show that pixel vignetting becomes more severe as CMOS technology scales, even for a 2-layer metal APS pixel. Finally, we briefly discuss several potential solutions to the pixel vignetting problem.

A new structure CCD sensor has been developed, which can be improved in the image quality compared with the conventional Inter-line Transfer type CCD sensors. The image sensor used in a high-definition image capturing system like a DS camera has been improved to have more number of pixels. This pixel increase adversely affects the sensitivity, dynamic range, signal-to-noise (S/N) ratios. The new CCD sensor has a new pixel arrangement of a half-pitch offset and a new structure of the transfer electrodes, and it achieves to increase sensitivity, dynamic range, S/N and the image quality. The new signal processing system combined with this new CCD has been also developed. It includes high quality analog and digital signal processing LSIs. The newly developed digital signal processing combined with over-sampling technologies enables to achieve higher resolution than conventional cameras with the same number of pixels. The outlines of the new CCD and the signal processing system, the configuration, operations, and the experimental results are described.

In spite the prodigious growth in the market for digital cameras, they have yet to displace film-based cameras in the consumer market. This is largely due to the high cost of photographic resolution sensors. One possible approach to producing a low cost, high resolution sensor is to linearly scan a masked low resolution sensor. Masking of the sensor elements allows transform domain imaging. Multiple displaced exposures of such a masked sensor permits the device to acquire a linear transform of a higher resolution representation of the image than that defined by the sensor element dimensions. Various approaches have been developed in the past along these lines, but they often suffer from poor sensitivity, difficulty in being adapted to a 2D sensor or spatially variable noise response. This paper presents an approach based on a new class of Hadamard masks--Uniform Noise Hadamard Masks--which has superior sensitivity to simple sampling approaches and retains the multiresolution capabilities of certain Hadamard matrices, while overcoming the non-uniform noise response problems of some simple Hadamard based masks.

Pixel design is a key part of image sensor design. After deciding on pixel architecture, a fundamental tradeoff is made to select pixel size. A small pixel size is desirable because it results in a smaller die size and/or higher spatial resolution; a large pixel size is desirable because it results in higher dynamic range and signal-to-noise ratio. Given these two ways to improve image quality and given a set of process and imaging constraints an optimal pixel size exists. It is difficult, however, to analytically determine the optimal pixel size, because the choice depends on many factors, including the sensor parameters, imaging optics and the human perception of image quality. This paper describes a methodology, using a camera simulator and image quality metrics, for determining the optimal pixel size. The methodology is demonstrated for APS implemented in CMOS processes down to 0.18 (mu) technology. For a typical 0.35 (mu) CMOS technology the optimal pixel size is found to be approximately 6.5 micrometers at fill factor of 30%. It is shown that the optimal pixel size scales with technology, btu at slower rate than the technology itself.

This paper describes the performance of an advanced high- resolution full-frame architecture CCD imaging device for use in scientific, medical and other high-performance monochromatic digital still imaging applications. Of particular interest is the replacement of the polysilicon 2nd gate electrode with that of a more spectrally transparent material thereby dramatically improving device sensitivity. This has been achieved without compromising performance in other areas such as dark current, noise, transfer efficiency and, most importantly, yield.

A digital programmable retina is a functional extension of a CMOS imager, in which every pixel is fitted with a tiny digital programmable processor. We actually call it a PAR, standing for Programmable Artificial Retina. From an architectural viewpoint, a PAR is a SIMD array processor with local optical input. A PAR is aimed at processing images on-site (where they are sensed) until they can be output from the array under concentrated form. The overall goal is to get compact, fast and inexpensive vision systems, e.g. for robotics applications. PAR design is subject to harsh constraints resulting from small pixel area and sensing/processing cohabitation. Meeting these constraints leads to using peculiar architectural and circuit technique solutions. In the last three generations of PARs we have designed, semi-static shift registers have played a crucial role in the maximization of computational power versus silicon area. In particular, the latter have been used to store, shift and--through some slight modifications--to perform local computations on images. Here, we show their abilities to support asynchronous propagation in order to implement `geodesic reconstruction', an extremely useful computational operator, in particular for image segmentation and then for object selection and manipulation purposes.

In this paper we describe a new technology which fabricates CCDs and fully depleted silicon on insulator CMOS circuits on the same 150-mm silicon wafer. We present results from 7.5 X 7.5-micrometers ² and 15 X 15-micrometers ²-pixel imagers that are 512 X 512 frame transfer devices. The 7.5-micrometers -pixel device exhibits a charge handling capacity in excess of 100,000 electrons at 3.3 V and the 15-micrometers - pixel device exhibits a charge-transfer efficiency over 99.998%. In addition, we demonstrate functional SOI CMOS ring oscillators with delay of 47 ps/stage at 3.3 V and 68 ps/stage at 2 V.

This paper presents results from a measurement program designed to investigate the variations in sensitivity of focal plane arrays on a sub-pixel scale. High-resolution measurements have previously been reported for a front- illuminated CCD device. New measurements are now provided on the sensitivity variations with in a single pixel of a backside-illuminated CCD. The measurements were made using a stable broadband light source and two high-precision translation stages. The pixel scans have been performed using 4 different filters (3 broadband and 1 narrowband). The results for each spectral region are presented.19

A novel method for the determination of a camera system's overall performance is described. This method involves the use of new form of test chart, termed a Nyquist chart. A Nyquist chart is a deterministic pattern of bright and dark structures that are imaged onto an image sensor array at the time sensor array's (spatial) Nyquist frequency. Camera system parameters such as MTF at the Nyquist frequency, distortion, pixel aspect ratio, color filter array layout, color balancing, field of view, relative illumination, depth of focus, and Petzval curvature can be readily determined.

A family of backside illuminated CCD imagers with 6.6 micrometers pixels has been developed. The imagers feature full 12 bit (> 4,000:1) dynamic range with measured noise floor of < 10 e RMS at 5 MHz clock rates, and measured full well capacity of > 50,000 e. The modulation transfer function performance is excellent, with measured MTF at Nyquist of 46% for 500 nm illumination. Three device types have been developed. The first device is a 1 K X 1 K full frame device with a single output port, which can be run as a 1 K X 512 frame transfer device. The second device is a 512 X 512 frame transfer device with a single output port. The third device is a 512 X 512 split frame transfer device with four output ports. All feature the high quantum efficiency afforded by backside illumination.

We have established our ability to deposit filters and coatings directly on delta-doped CCDs by constructing a metal/insulator molecular-beam epitaxy (MBE) chamber connected via an ultra-high-vacuum transfer line to the silicon MBE chamber, in which delta-doping is performed. We have fabricated preliminary MgF₂ AR coatings on delta- doped SITe CCDs.

A versatile post-fabrication process to produce thinned, flat, back-illuminated charge-coupled devices (CCDs) has been developed at Jet Propulsion Laboratory's Microdevices Laboratory. This technique is compatible with many ultraviolet enhancement treatments and has been demonstrated with the delta doping process. The significance of this demonstration is that thinned, robust, and flat CCDs are produced without the use of epoxies or waxes using temperatures and materials that are compatible with standard CCD fabrication and delta doping processes. In our approach, the CCD is attached by thermocompression bonding to a specially-designed silicon substrate using gold-gold diffusion bonding prior to thinning. CCDs with optically flat membranes (10 - 20 micrometers ) were produced with excellent yield. These flat CCDs have been successfully delta doped. We will discuss the process of producing thinned flat CCDs, their delta doping, and our results to date.