Improved Single-Chip Camera Systems for Multimedia Applications
Advances in CMOS image sensors, and their associated commercialisation, have led to the need for improved modelling of the photodiode sensing element. The goal of this research is to develop accurate high- and low-level design tools for CMOS active pixel sensors. Objectives of the project include:
- Modelling of individual photodiodes: uniformity, dark current, and optical response in terms of process parameters and diode geometry.
- Array Modelling: Modification of above analyses to incorporate the effects of neighbouring photodiodes in an array
- Optical Analysis of Active Pixel Sensor: including modelling of overlayers (passivation, metal, etc.)
Adaptive-Scanning Single-Chip Cameras for Industrial Inspection
Industrial inspection by machine vision requires image capture at high scan speeds and high resolution. In some applications, however, only a portion of the imaging area is of interest at any one time, so significant improvements in throughput can be achieved by reducing the quantity of data transmitted by the camera. A novel system is under development here which combines both greyscale image and motion detection in the same array of pixels. The motion information will be used on-chip to define the size and/or location of a selectively scanned region (e.g. captured at a higher resolution). The selected area will therefore adapt to, and "travel" with, the moving object. This is achieved on a single chip by employing the standard CMOS fabrication process, which permits the integration of the sensor with its control, processing and output electronics. The result will be a reduced off-chip data rate and a compact, robust and inexpensive single-chip sensor system for industrial, communications, and security applications.
Materials and Manufacturing Ontario
Integrated detector systems, such as those based on micro-bolometers or optical, open up the possibility for the first time of combining hundreds or thousands of individual sensors into a single mosaic array. Individual elements of the array may have different spectral sensitivities, different fields of view, or be physically remote from one another. Hitherto, sensor cost and inter-sensor communication bandwidths for such a system have been prohibitive, but on-sensor integration techniques developed during this project will enable the practical demonstration of a prototype mosaic imaging system. This is to be achieved by designing each chipxel of the array such that it only communicates with the rest of the system when it detects a pre-determined image, or icon, of interest.
While the specific application of the mosaic concept is for micro-satellite based infrared imaging for forest fire detection, potential applications for the technology are widespread and the potential influence is profound. Multi-camera systems are employed in industrial inspection (particularly for web inspection of rolls of cloth, paper, etc.). Currently, the maximum number of modules is six to eight and this application would clearly benefit from the technology to be developed in this project. Other potential applications include security systems in which many cameras must be co-ordinated, especially if the link between them is slow or physically long. Here too, the proposed technique of matching icons representing events of interest would be applicable.
Sponsored by:
Centre for Research in Earth and Space Technology
Topaz Technology Inc.
Algorithm Development for the Spectral Sun Sensor
The Spectral Sun Sensor consists of an electronic image sensor, onto which an image of the sun is projected through a dome containing twelve pinholes. A CMOS active pixel sensor is proposed for this application, owing to their low-cost, high levels of integration and radiation tolerance. From the shape and position of the projected image and knowledge of the pinhole from which the image originates, the Sun’s position can be determined.
Hence the objective of the proposed project is to develop both off- and on-chip processing concepts and appropriate algorithms for the Spectral sun sensor to enable:
1) the efficient computation of the centroid of the spot projected by the Sun onto the detector array, and 2) the determination of the specific pinhole associated with the projected spot.
High-Resolution Artificial Retina
This research project aims to construct an electronic imitation of the human eye, for use in practical engineering applications such as industrial inspection, communications, and security. Conventional machine vision systems comprise a camera and a computer in a way that is analogous to the human eye and brain. But the eye is much more than just a sensor. It also carries out image processing functions on a pixel-to-pixel level, which greatly enhance the performance of the eye/brain system. One of these functions, for example, significantly increases the range of light levels (maximum and minimum brightness) that the eye can detect. So, can we make an artificial camera that acts the same way? While some basic functions of the eye can be imitated electronically, no practically useful artificial retina has yet been made. The main reason for this is that the eye is made up of layers of cells, each performing a specific function (light detection, image processing, communication, etc.), whereas a silicon chip is essentially two-dimensional. So any advanced functions in a chip must be spread out sideways over the surface. This leads to large intervals between the optical sensors, and the image resolution is unacceptably low.
In this project, we are developing methods for achieving a high resolution artificial retina sensor, and will demonstrate prototypes of these chips for practical engineering applications. Most artificial retinas investigated to-date sacrifice resolution for accurate implementation of a biological model. We will take the alternate approach of making acceptable approximations to the model in order to achieve a high image resolution. For example, we can partition the camera sensor and the image processing area into two separate regions of the chip, so that the processing and imaging operations do not interfere. Alternatively, we can process only selected pixels, thereby reducing the chip area needed for the processing operations and increasing the resolution. Each of these options leads to a number of design trade-offs, which will be investigated and characterised. This work relates closely to other industrially relevant research in our laboratory on integrated smart camera systems.
Sponsored by:
Pixelated CMOS Photodiode
o image acquisition at multiple resolutions
o straightforward region of interest identification and fast scanning of that region
o graceful radiation-induced degradation
o facilitation of intelligent power control
o incorporation of different pixel geometries
o system-on-a-chip integration with direct computer interface
The intermeshed array technology is ideally suited to applications where the important signal covers a relatively small proportion of the optically sensitive area, such as lidar and optical intersatellite links, for which the region-of-interest readout may dramatically improve the signal to noise ratio. Compared to conventional pin junction photodiodes, the CMOS photodiode is expecte to have an order of magnitude improvement in sensitivity for the above applications. Besides an increased SNR, the CMOS photodiode will integrate all support circuits into a single low-power chip with a convenient serial computer interface. The ‘intermesh’ principle allows the average power dissipation to be reduced from the already-low levels of typical CMOS imagers without sacrificing other performance features.
For space missions, the intrinsic megarad tolerance of new CMOS fabrication technology is augmented by the graceful degradation under single-event upset that is possible with the ‘intermesh’ concept. Such a component allows missions into high-radiation environments like the Van Allen belts.
Topaz Technology Inc.
An estimated five million North Americans suffer from low vision, the general condition of reduced or impaired vision without full blindness. This number is expected to increase as the proportion of elderly people in the population grows. People with low vision can benefit from a number of aids, such as telescopes, closed-circuit TVs, and other devices which modify visual information according to patients' needs. Most of these devices are bulky and static, so a major improvement in quality of life can be achieved by making a wearable head-mounted system; cameras record the person's surroundings, appropriate image processing is performed, and the modified image is presented to the patient via displays for each eye. However, current low vision enhancement systems are bulky, complex to operate, difficult to adjust for each patient, high power, and unattractive.
To achieve the goal of a vision enhancement system which will fit into a device similar to a regular pair of glasses, major improvements are needed in the electronic cameras. They must be highly integrated, compact, lightweight, low power, and highly automated. This project concentrates on the development of a suitable electronic camera system that would integrate much of the image processing functionality onto the sensor chip. Example functions include automatic contrast enhancement, digital zoom (not requiring zoom lenses), automatic brightness adaptation, wide dynamic range, and high sensitivity. Programmability to the patient's specific needs is also required. No such camera is currently available. The implementation and optimization of each of these functions involves significant technological innovation.
This project is an inter-disciplinary collaboration combining the electronic camera design expertise in the University of Waterloo Integrated Camera Group, the clinical experience with low vision patients at the School of Optometry, and the commercial expertise of Betacom, a recognized innovator in life-enhancement technologies.
Premiers Research Excellence Award
Article on this work from Kitchener-Waterloo Record,
6 October 2000 (large file)
Article from L'actualite (in French), October 2002
Radiation Effects in Deep Submicron CMOS Image Sensors
Topaz Technology Inc.
The objective of the proposed project is to establish an experimentally verified theoretical framework for understanding radiation effects in deep-submicron CMOS integrated circuits (ICs), and image sensors in particular. The emphasis will be on the derivation of engineering formalisms to assist the design process.
MIKE
Currently MIKE identifies geometrical shapes using a binary thresholding approach. It is capable of around 24 saccades per second, several times faster than the human eye.
