We are developing a retinal prosthesis vision chip based on a pulse-frequency-modulation (PFM) photosensor. Because the device is implanted in the eye ball and is powered by RF coil coupling with limited energy, low voltage and small current operation is required to achieve low power dissipation. We propose a capacitive-feedback-reset method for the PFM vision chip. The proposed method uses capacitive feed back through the junction capacitance of the photodiode and gate-source overlap capacitance of the reset transistor. In the proposed PFM circuit, the feed-through effect in resetting contributes to avoid current competition, so that the high dynamic range can be achieved even at the low voltage operation. We have fabricated a pixel TEG circuit in a 0.35-μm CMOS technology. The PFM photosensor circuit is composed of a four-stage inverter-chain. Dynamic range of 136dB has been achieved with 0.8-V power supply.

Image sensors with pulse modulation measurement scheme are fabricated for bioimaging and biosensing ap-plications. We designed pulse modulation photosensors, a 64×64-pixels image sensor for in vitro bioimaging, and a 176×144-pixels (QCIF) image sensor for in vivo bioimaging. We demonstrated the feasibility of the pulse modulation measurement scheme for biosensing applications. We obtained a dynamic range of 120dB and minimum sensing intensity level of 2nW/cm². We also confirmed that 0.2% of intensity change is detectable at the minimum intensity region. An in vitro, on-chip imaging of a mouse hippocampus was successfully demonstrated. A sensor module for in vivo imaging is also developed.

Color imaging systems still use a combination of conventional photo-detectors and RGB optical filters for color measurement. This entails many limitations to the color sensor performances. We reported that buried junction color detectors give good alternatives to overcome these limitations. However, successful design of color sensors using these detectors requires an accurate knowledge of their behaviors. Unfortunately, circuit simulators do not provide models for these devices. In order to make the designer task more flexible, an optoelectronic library is developed under CADENCE design tool. It consists of some optoelectronic elements such as Buried Double pn Junction (BDJ) detector and a set of optical sources. This allows the designer to choose an optical stimulus with a specific spectral distribution and also to select the total power incident on the BDJ surface. The library is obtained by implementing, in Spectre simulator, the behavioral models of the optoelectronic elements. The models are written using Verilog-A language. Simulations of the BDJ spectral response and dark currents give a good agreement with experimental data. We also note the absence of convergence errors or mathematical faults during DC and transient simulations of active pixel sensor architectures. These results confirm the robustness of the optoelectronic library.

We developed an ultrahigh-sensitivity camera tube with a 15-μm-thick high-gain avalanche rushing amorphous photoconductor (HARP) film and applied it to an HDTV camera. The camera, called the "New Super-HARP", can achieve about 30 times the sensitivity (62.5 lux, F10) of conventional HDTV CCD cameras. Furthermore, for slow-moving subjects, the camera can dramatically increase the sensitivity in the intermittent read-out mode (for an accumulation time of 4 seconds, about 240 times the sensitivity of a New Super-HARP camera under normal operations). The very-low-dark-current feature of the HARP film results in excellent video images without any fixed pattern noise. We investigated the relationship between the operating temperature of the film and the occurrence of highlight defects in 15-μm-thick HARP films when shooting fixed, strong spot-lights directly. We found that defects could be suppressed by shifting the operating temperature to 35°C from the conventional 25°C. Furthermore, we optimized the concentration of arsenic (As) doped in the film to improve the heat resistance so that the film could be used at temperatures as high as 35°C. Ultrahigh-sensitivity imaging technology using HARP has been attracting considerable interest from many fields outside of broadcasting, such as medicine, biology, and digital film production.

Programmable artificial retinas (PAR) are image sensors with a digital processor in each pixel. PAR based vision systems are our framework. In its basic version, a PAR operates in SIMD mode, which restricts it to low level vision only. To support higher levels of vision, non-SIMD operators have to be settled in each pixel. Programmable connections and asynchronous communications are key ingredients to support regional computations, but under which form? To identify the good ones, we consider adding pixel data simultaneously over different regions as an exemplary primitive. After recalling previous implementations, we propose a novel one based on a tree extension of micro-pipelines that we call convergent micro-pipelines.

A 10T/pixel CMOS active pixel sensor with clock count output, ultra low supply voltage, and wide dynamic range is presented. This pixel comprises a reset transistor, photo-diode, a comparator, and an inverter with pixel size of 9.4x9.4μm² and 24% fill factor in a standard 0.25-μm CMOS logic technology. The output transfer curve is the same as an amplified logarithmic-response and is similar to the light response of the human eye. Besides, the pixel can operate at an ultra low supply voltage and the output characteristics will not be affected. Even with supply voltage down to 1.2V, the dynamic range of this pixel can remain as high as 90dB.

A 32x32 visual motion sensor based on the time stamped pixel structure is proposed. The sensor pixel is formed by a time stamp component and a compact spatial based edge detector with a pixel size of 70x70μm² in standard 0.35μm CMOS process. It measures up to 6000 degree/s with focal length f = 10mm. Less than 5% rms variation for middle range velocity measurement and less than 10% rms variation for high and low velocity measurement have been achieved. This structure is good for scaling down with new fabrication processes to implement large scale 2D arrays with low power consumption in the near future.

We describe a CMOS image sensor with pixel level analog to digital conversion (ADC) having high dynamic range (>100db) and the capability of performing many image processing functions at the pixel level during image capture. The sensor has a 102x98 pixel array and is implemented in a 0.18um CMOS process technology. Each pixel is 15.5um x15.5um with 15% fill factor and is comprised of a comparator, two 10 bit memory registers and control logic. A digital to analog converter and system processor are located off-chip. The photodetector produces a photocurrent yielding a photo-voltage proportional to the impinging light intensity. Once the photo-voltage is less than a predetermined global reference voltage; a global code value is latched into the pixel data buffer. This process prevents voltage saturation resulting in high dynamic range imaging. Upon completion of image capture, a digital representation of the image exists at the pixel array, thereby, allowing image data to be accessed in a parallel fashion from the focal plane array. It is demonstrated that by appropriate variation of the global reference voltage with time, it is possible to perform, during image capture, thresholding and image enhancement operations, such as, contrast stretching in a parallel manner.

As digital imaging arrays increase in size and resolution, defect correction could lower costs and improve yields. A fault tolerant active pixel sensor (APS) has been designed that will operate in the presence of a single point defect. The photosensitive area of the pixel is split in half and both halves operate in parallel. The output of each half is combined using a common row select transistor. The common pixel defects are optically stuck high (bright pixel) and optically stuck low (dark pixel). Simulations showed that a non-defective pixel would function normally and if one pixel half was defective, the other half would operate normally with half the sensitivity of a non-defective pixel. Fault tolerant photodiode and photogate APS’ were designed and fabricated in CMOS 0.18-micron technology. Half stuck high and half stuck low defects were induced on the fault tolerant pixels and the sensitivity ratio of non-defective to half defective pixels was measured (ideally 2). The experimental ratios ranged from 1.89 (stuck high) and 2.02 (stuck low) for the photodiode APS to 1.73 (stuck low) and 1.77 (stuck high) for the photogate APS. Non-defective fault tolerant pixels have also shown a 2x increase in sensitivity over normal APS’.

We propose a CMOS image sensor which can capture both whole images at 60 fps and several region-of-interests (ROIs) images at a high-speed frame rate of 2.1 kfps. We have designed a sensor prototype for our newly developed man-machine interface system called “Opto-Navigation”. In the system, LEDs as optical beacons are equipped with portable devices such as mobile phones and digital cameras, and provide a useful guide for visually aided data communication with them. The optical beacons inform their positions at slow frequency around 10 Hz, which define ROIs, and send their ID data with bit rate of 1kbps to the sensor. The sensor can recognize the position of IDs with frame differences of 60-fps images and then obtain the ID data by reading only ROIs at high frame rate. Thus by using the image sensor a user can know what devices can be communicated with and select one of them to send/receive information data. A QVGA image sensor dedicated for the “Opto-Navi” system has been fabricated in a 0.35-μm 2-poly 3-metal standard CMOS technology. The pixel has 7.5 × 7.5 μm² area with a fill factor of 24.9%.

The Modulation Transfer Function is a common metric used to quantify image quality but inter-pixel crosstalk analysis is also of interest. Because of an important number of parameters influencing MTF, its analytical calculation and crosstalk predetermination are not an easy task for a CMOS image sensor, due to the use of several metal line and transistor in a close proximity of the photodetector. A dedicated test chip (using a technology optimized for imaging applications) has been developed in order to get both MTF data and influence of the various areas of the pixel to its own response and the one of its neighbors. In order to evaluate the contribution of pixel elementary patterns (particularly the in-pixel readout circuitry), several kernels of shielded pixels have been implemented with the central pixel locally unmasked. Analyze of the kernel responses provides a good insight on both Quantum Efficiency and crosstalk contributors. Additionally, the top metal layer has been used to implement metal edge pattern allowing the on-chip measurement of Edge Spread Function so the MTF. The results obtained with pixel kernels and direct MTF measurements, performed on the same chip at different wavelengths, are analyzed and compared in order to correlate them and draw conclusions that can be applied at the design level.

In this paper, we describe the design and implementation of a one-chip camera device for a capsule endoscope. This experimental chip integrates peripheral circuits required for the capsule endoscope and the wireless transmission function based on a data transmission method using human body conduction. The integrated functional blocks include an image array, a timing generator, a clock generator, a voltage regulator, a 10b cyclic A/D converter, and a BPSK modulator. It can be operated autonomously with 3 pins (VDD, GND, and DATAOUT). A prototype chip which has 320x240 effective pixels was fabricated using 0.25μm CMOS image sensor process and the autonomous imaging was demonstrated. The chip size is 4.84mmx4.34mm. With a 2.0 V power supply, the analog part consumes 950µW and the total power consumption at 6 fps (20MHz carrier frequency) is about 3mW.

Zernike moments are operators that are often used in the field of image analysis and pattern recognition. A certain number of retina chips that implement geometrical moment function have already beeen described in the litterature. However, the architectures of these circuits are not programmable so that their field of application is limited to position detection of an object and they are not suitable for pattern recognition. In our paper, we propose a method and its implementation in a programmable retina circuit that allows us to compute, with the same circuit, Zernike moment values of different orders and repetitions. Our method is based on the measurement of the correlation value between two binary masks (memorized in memory devices integrated at the pixel level on the sensor) and an image to analyze, projected onto the sensor by optical means Indeed, if we consider the real and imaginary parts of the Zernick polynomial of order p and repetition q as two images, then we can notice that there is a close relationship between the correlation value of two images and the expression of the real and imaginary parts of the Zernick moments of an image. Thus, the value of the Zernick moment of an image can be obtained by computing the correlation value between the image under analysis and two other images, one for the real part and another one for the imaginary part. The latter two images that depend on the order p and repetition q of the Zernick moment to compute are gray level images that need to be memorized in the retina. In order to reduce hardware implementation cost they are transformed into binary images or masks using a dithering algorithm. In this way only a 2-bit memory device is required per pixel to memorize the two masks (on bit per mask). Using the binary masks instead of the gray level images only gives an approximate value of the Zernick moments. However, we will show that the approximated values are still a good representation of the analyzed image (and thus can be used in a pattern recognition application). To do so, the exact and approximate values of the Zernick moments for values of p and q ranging from 0 to 30 have been computed and the images reconstructed from these values compared to the original one. The relative errors between the respective reconstructed images (exact and approximated Zernick moments) and the original image have been plotted against the orders of the Zernick moments used in the reconstruction. We have noticed that the evolutions of the error curves are quite similar. In the final paper we will also present the architecture of the CMOS retina circuit that implements the Zernike moment computation function as well as the simulation results. The circuit has been designed in standard 0.35µm CMOS technology and it is composed of an array of 180 x 180 pixels.

The advent of high speed, CCD-based camera technologies opens new possibilities for field monitoring applications. In particular, under natural or man-made loading conditions, applying these new technologies towards the monitoring of building interiors may substantially help rescue and reconnaissance crews during post-event evaluations. To test such a methodology, we have developed a specialized network of high-speed cameras and supporting hardware for monitoring and tracking nonstructural elements subjected to vibration loading, within building structures. Teamed with the University of California, Los Angeles, a full-scale vibration experiment is conducted on a vacant structure damaged during the 1994 Northridge Earthquake. The building of interest is a four-story office building located in Sherman Oaks, California. The investigation has two primary objectives: (1) to characterize the seismic response of an important class of equipment and building contents and (2) to study the applicability of tracking the response of these equipment and contents using arrays of image-based monitoring systems. In this paper, we describe the image acquisition (hardware and software) system and the experimental field set-up are described. In addition, the underlying communication, networking and synchronization of the camera sensor system are discussed.

A concept and preliminary design of an Extreme Ultraviolet (EUV) spectrometer is presented. The spectrometer is based on a gas ionization chamber and an advanced eight-electrode electron focusing system to form a narrow electron beam on a photodiode aperture. The design is modeled with the SIMulation of IONs (SIMION) tools and shows the ability to scan through the spectral range of 20.0 - 40.6 nm by changing the potential on a single control electrode between about 200 and 1100 V. The spectral resolution is about 0.25 nm in the middle of the band. The set of the focusing potentials may be changed to allow detection of solar EUV radiation in a wider spectral band, e.g. 5.0 - 50.0 nm. The potentials may be also optimized to improve the spectral resolution in a required spectral window.

A 1/2 inch 1M-pixel monochrome Frame-Transfer CCD imager with 5.6μm by 5.6μm pixel size was developed for use in medical and industrial applications. The sensor production uses 17 mask steps designed for an improved process, with highly transparent membrane poly-silicon gates and two metal layers. The first metal layer is used for vertical strapping to reduce the RC- times of the imaging electrodes. The image pixels, the storage cells and the readout register are made using two layers of membrane poly-silicon. An n-channel implant on a profiled p-well in a n-substrate achieves 52,000 electrons full well charge storage capacity in combination with excellent vertical anti-blooming and fast electronic shuttering. Smear as low as 0.06% at 1/30 sec integration time is achieved at 5MHz frame shift frequency. The pixel charge is converted to an output voltage using a 3-stage source follower amplifier, optimized for 40MHz pixel frequency. For use in high-speed industrial applications, the split read-out allows pixel rates up to 80MHz. The output amplifier with a conversion gain of 18.7μV/electron has an rms noise of 18 electrons at full bandwidth (linear dynamic range of 67.8 dB). The dark current level is 100 pA/cm² at 60° C.

The paper describes important design features and resulting performance of a color VGA format Frame Interline Transfer CCD image sensor that utilizes Charge Carrier Multiplication for increased sensitivity and low noise. The description includes the details of the photo site design that is formed by a pinned photodiode with a lateral anti-blooming drain. The design details of the photo-site transfer gate region are also given together with the design details of the vertical CCD register that result in a fast charge transfer into the memory and thus low smear. Since the device is not using the vertical overflow anti-blooming drain for the booming control, the near IR performance is not reduced. The color sensing capability is achieved by employing either RGB or complementary color filters. The remaining focus of the article is on the typical characterization results such as the CTE, image lag, and low dark current. The NIR and color imaging performance at low light levels is investigated in detail and characterized. In conclusion several typical scene color images taken by the camera that uses the developed charge-multiplying FIT CCD image sensor are shown.

CCD is a continuum of MOS capacitors, so its big capacitance becomes one of the major disadvantages compared with CMOS image sensor, that cause not only large power dissipation but also other problems, such as generating an electro magnetic interference. We have developed a CCD linear image sensor with thin single-layer electrodes for the purpose of reducing the CCD capacitance. A two phase pulse drive CCD is fabricated with single layer poly Si electrode that has narrow electrode gaps and thinner electrode thickness. At the sensor that has 2.625um pitch 10k pixel linear array with a single sided CCD register, the coupling capacitance has been reduced to totally less than 40% compared to the conventional two layer CCD electrode structure, due to non electrode overlapping and thin thickness of the CCD electrodes. The total power consumption for CCD drive is reduced to 45% of conventional CCD and high transfer efficiency (>99%) is obtained at 20MHz. Moreover, the size of the area around CCD for the contact between electrode and clock applying wire is reduced by eliminating second layer electrode. The flatness above the silicon surface is also improved for better image quality.

Life recognition is desired for unattended fingerprint identification. We look at the color changes in a series of fingerprint images acquired during the course of an input action. As we press a finger upon an input device, a fingerprint area gradually increases and its color changes. This is due to the blood movements induced by a finger deformation. For example, we can declare a finger is alive if its color change exceeds a certain value. However, the color changes in the images acquired by a conventional device using a prism are very small. This is because the light reflected by the valley regions of a finger carries little information about the blood movements. We tried a fingerprint sensor based on scattered light detection. Here, a plastic plate serves as a light-guide for LEDs mounted on its edge. When a finger is pressed against the plate, the propagating light is scattered by the finger and leaks out of the plate. A standard color CCD camera placed near the plate captures the scattered light. Data extracted from more than ten people of different age groups indicate that there is a certain minimum value for the color changes.

A low cost, compact electronic image leveler using a CMOS image sensor is developed for level measurement. Optical system with two lenses is designed to get tilt information of total system. One lens is designed in bowl shape that can hold liquid. The gravity of the light spot on CMOS image sensor is converted into pixel coordinates proportional to the tilt angle to be measured. The reading tilt angle could be estimated with resolution better than 10 arc second.

This paper describes an Airborne Multi-Spectral Imaging System (AMIS) and the development of its system software. This system has been developed so as to be rapidly deployed in response to episodic events such as hurricanes and tropical storms which may occur year round in coastal zones. The system uses digital video cameras to provide high resolution images at a very high collection rate. The system is software controlled so as to provide a minimum distraction for the aircraft pilot by providing for the remote manipulation of the camera and the GPS receiver. The system is viable for many applications that require good resolution at low cost. Such applications include vegetation detection, oceanography, marine biology, and environmental coastal science analysis.

With a band gap of silicon of 1.1eV, the largest wavelength that can excite electrons from the valence to the conduction band is roughly 1100nm. As a consequence, in, for instance, a charge-coupled device, the quantum efficiency (QE) for wavelengths larger than 1100nm is assumed to be zero. We found that there is a response at those longer wavelengths and that the response decreases with increasing wavelength. The QE increases with increasing chip temperature which suggests a thermally activated process. Impurities in the silicon provide the energy levels in the band gap, from which electrons can be excited either thermally or by absorption of a photon. It is these impurities that contribute to the infrared response. We characterized the response at chip temperatures of 248 K to 293 K for wavelengths from 1200 nm to 1600 nm and calculated the activation energies at these wavelengths. We found that hot pixels, i.e., pixels with extraordinary high counts in a dark frame, tend to respond stronger to infrared light than normal pixels. This correlation gets stronger for longer wavelengths. It is argued that this response can be used for probing the impurities present in the silicon bulk of the sensors.