Data processing: measuring – calibrating – or testing – Measurement system – Measured signal processing
Reexamination Certificate
1998-07-07
2001-12-25
Hoff, Marc S. (Department: 2857)
Data processing: measuring, calibrating, or testing
Measurement system
Measured signal processing
C250S208100
Reexamination Certificate
active
06334098
ABSTRACT:
This is a national phase application claiming priority from PCT/IL97/00356, filed Nov. 6, 1997, which, in turn, claims priority from IL 119596, filed Nov. 11, 1996.
FIELD OF THE INVENTION
The present invention relates generally to electronic sensors and cameras that acquire images, such as video cameras in the visible and invisible spectra of light. More specifically, the present invention relates to electronic circuitry and arrangements that perform in real time in order to increase the dynamic range of electronic image sensors and cameras.
BACKGROUND OF THE INVENTION
The Dynamic Range (DR) of images of scenarios and views is defined as the ratio between the intensity of the brightest spot in the image and the minimum significant differential of intensities across the image. The image DR determines how many distinct levels of intensities a subsequent user of an image must employ in order to fully exploit the information conveyed by the image. In sensor acquired images—the minimum significant cross-image difference is bounded by the sensor noise, and on the other hand—the maximum acquirable signal is bounded by the sensor saturation level. One can thus realize that the DR of an acquired image is in fact bounded by the Saturation to Noise Ratio (SatNR) of the acquiring sensor. The DR of natural views and scenarios may exceed ten to the tenth power ( 10
10
). The human eye, as well as the eyes of animals, can accommodate such huge DR without too much of an apparent difficulty. This implies that the self generated noise of biological image sensors is sufficiently small compared to their saturation level, so that the SatNR can match the DR of natural views and scenarios. However, the SatNR of common electronic image sensors such as Charged Coupled Devices (CCD) does not usually exceed several hundreds, say 256. This in turn implies that all the details in the original view whose contrasts are smaller than {fraction (1/256)} of the brightest spot, but still are bigger than say, ten to the minus tenth of the same, will totally disappear in the process of acquisition.
Thus, it is the purpose of many inventions including the present, to increase the DR of image sensors in order to diminish the acquisition losses as much as possible. The inherent sensor sensitivity, which is how much of electrical charge is generated by the sensor per a given number of photons that hit its receiving surface, depends on its geometric and optical properties and the quality of the materials that make it. Traditionally, the essence of the imaging process, which is to say—the conversion of images from levels of light intensities into levels of readable electrical charges or voltages—is the exposure of an appropriate array of sensors to an adequately focused image for a prescribed period of time, letting those sensors collect photons and convert them into electrical charges. With time those charges are accumulated, so that at the end of the exposure interval the amount of the net accumulated charge at a particular sensor reflects the magnitude of the photon flux, or the light intensity that belongs to the spot of the image that has been focused on that particular sensor. An obvious way to influence the sensors responsivity to light flux, is varying the exposure time interval, also referred to as “The Integration Time”. Another obvious such way is to change the sensor optical aperture. Increasing the exposure time and/or the aperture enable the sensors to integrate more photons at every spot in the field of view, hence causes a proportional increase of the accumulated charges. However, besides other side effects of overly increased exposure interval and aperture size, there is a maximum amount of charge that a light cell can hold. When this maximum level is reached in a certain cell, it becomes saturated and ceases to reflect the light variations of the original view. At this point this cell can not function as an imaging device anymore. It is commonly said that the whole image is in saturation when considerable areas of neighboring cells are saturated, causing these areas of the output image to appear as white uniform zones that do not show any of the original view. The only way to increase the DR of an image sensor of a given sensitivity, is to increase its maximum Signal to Noise Ratio (SNR). The maximum SNR of a sensor is the sensor Saturation to Noise Ratio (SatNR). Thus, in order to increase the DR of a sensor one can either try to increase its saturation level, or one can try to reduce the sensor self generated noise. There are two principle kinds of self generated noise:
a). Temporal noise, a stochastic process that depends mainly on the size and temperature of the active part of the sensor, and
b). A so called “Fixed Pattern Noise” that is associated with sensors that consist of arrays of image detectors, also called “Light Cells”. Most of the fixed pattern noise is originated in the initial, or bias voltage that a light cell may have prior to its exposure to light. In arrays of sensors those bias voltages may vary from one sensor to another, so that the exposure of the array to an absolutely uniform, so called “Flat Field” input image, will still result in a non uniform output image that merely reflects the variations in the sensors bias terms. Reduction of temporal noise is done by reducing either, or both the sensor size and its operating temperature. Reduction of the fixed pattern noise is done traditionally by improving the uniformity between the light cells inside the array, and by using various methods of subtraction or compensation of the fixed pattern, or as it also called, the flat field image.
THE PRIOR ART
Temporal noise reduction by cooling is done in CCD cameras made by many manufacturers. One example is Reticon Corp. in U.S.A. Fixed pattern compensation is done in real time hardware by CID Corp., U.S.A. Flat field subtraction is done in real time software by Coreco Corp., Canada. I-Sight Corp. from Israel has introduced the multiple exposure technique in order to increase CCD sensors DR by increasing their maximum acquirable light level. In this method a saturated zone from a longer exposure is replaced by its replica from a shorter one. It can be shown however, that the maximum possible gain of DR in this method is in a factor of two, or a single bit in a binary/digital representation.
SUMMARY OF THE INVENTION
According to the present invention, the photon flux that reaches each photodetector in an array of sensors is estimated in real time via an optimal estimation logic, also known as Kalman Filter. The essence of the optimal estimation concept is that the desired estimate of the photon flux is obtained via a weighted combination of several time derivatives of the sensors outputs, and not just the zero'th derivative, as is done traditionally. In this combination of derivatives the first derivative plays a major role. The estimation processes are performed individually and simultaneously for all the light cells. They start together at the beginning of the exposure interval, but when the outputs of individual light cells are nearing saturation, their associated estimation processes are automatically terminated. The present estimators mainly sense the first derivative, or the slope of the signal at each light cell output. This slope is on one hand a measure of the light flux that reaches a detector, and on the other hand its value is independent of the exposure time, hence it is totally isolated from the effect of saturation. The integration time in this method influences only the SNR of the estimated result. Thus, using for example, the longest possible exposure and the largest available aperture, the present invention enables one to exploit the maximum achievable SNR of a particular sensor, and at the same time to avoid the saturation effect altogether. Also, the fact that the output of the present device represents mainly the slope of the sensors output, effectively eliminates the fixed pattern noise from the acquired image. Thus, the present invention makes the DR of a sensor
Friedman Mark M.
Hoff Marc S.
Raymond Edward
Truesight Ltd.
LandOfFree
Increasing the dynamic range of image sensors does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Increasing the dynamic range of image sensors, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Increasing the dynamic range of image sensors will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2559827