Adaptive Sensitivity™


Adaptive Sensitivity is a visual information processing feature that mimics the human eye's ability to compensate for uneven illumination in high-contrast scenes, delivering an optimally contrasted image without distortion or loss of quality.

How it works

Input to the camera’s DSP are two simultaneous mosaic format digital video signals. The two input signals are referred to as "Long" and "Short".
The "Short" video input should contain data from the input image as acquired by a CCD with a short exposure time. The bright regions should register faithfully, while the dark regions may be close to or at cut-off, and hence meaningless.
The "Long" video input should contain image data as acquired by a CCD with a long exposure time. The bright regions may be at saturation, but the darker regions should register faithfully.
The Adaptive Sensitivity algorithm for wide dynamic range video imaging combines the meaningful information from both “Short” and “Long” inputs, and generates a single output image in which the luminance range is compressed to video standards (48 dB or 8 bits), without loss of image detail or color fidelity.

More informtion

The sensor's output signal is separated to Long and Short signals, pre- amplified, sampled by a correlated double sampler and then digitized and fed into an iSP2000 DSP. The input image data from each exposure is separated into the luminance ("luma") and chrominance ("chroma") information. The "Long" and "Short" luma and chroma data are pre-processed and combined by a weighted summing operation.
Internal look-up tables adjust the summing weights and control the picture contrast and brightness.
Finally, the processed components are transformed into digital Y/Cr/Cb, Y/I/Q, R/G/B, NTSC/YC and 601/4:2:2 formats (software selectable) or to analog signals.
In the monochrome configuration, only the luminance channel and the Y output are used.
In addition to its processing functions, the iSP2000 also provides centralized digital control over camera functions.
It internally collects luminance, edge and color statistics for auto-exposure, auto-focus and white-balance functions.
It is possible to assign different statistical weights to different regions within the frame.
For example, an auto-exposure application which accumulates the average luma of a given frame, might weigh the center of the picture at 1, the bottom at ½ and the top at ¼.

Demo Images

The figures below provide an example of Adaptive Sensitivity™ principles. The 'Long Exposure' and 'Short Exposure' Figures show images of a scene requiring imaging with wide dynamic range. Each image lacks the details which may be seen in its counterpart. The central image, which is an image produced by the Adaptive Sensitivity™ algorithm, clearly illustrates the benefits offered by both exposures.

 

Long Exposure Adaptive Sensitivity Short Exposute

Long Exposure Adaptive Sensitivity Short Exposute

Long Exposure Adaptive Sensitivity Short Exposute

Long Exposure Adaptive Sensitivity Short Exposute

Demo Movies

Problem Statement

The (intraframe) dynamic range of a camera is usually defined as the ratio of the brightest point of an image to the darkest point of the same image. It is also called the maximum contrast of that image. Unfortunately, the dynamic range of most electronic cameras is severely limited. It is narrower than the dynamic range of most scenes, and it is also more limited than that of photographic film.

Imaging applications often deal with situations in which lighting conditions are far from optimal. In particular, these may include objects positioned against strong back lighting, in which case the objects details become too dark, since the camera adjusts itself to the high average brightness. In some situations there will be many spots with steep gradations of brightness, which are hard to handle by standard cameras. Other situations depend on the dynamic behavior of the camera: abrupt changes of illumination will cause profound transition effects on the overall system.

All the above situations call for wide dynamic range imaging, which is generally constrained by three factors: the sensor, the signal processing circuits, and the display (or frame grabber).

Common CCD sensors can acquire a contrast of roughly 1:1000 (60 dB) dynamic range of intensities. The darkest signal is constrained by the thermal noise, or "dark current", of the sensor. The brightest signal is limited by the total amount of charge that can be accumulated in a single pixel. Usually CCD chips are built such that this total maximum charge is about 1000 times the charge generated thermally. This dynamic range can be substantially enhanced, in special applications such as scientific or astronomical cameras for still imaging, by cooling the sensor and by employing special readout circuits. However, such methods, in addition to being very expensive, are inappropriate for real-time applications.

As described above, many applications usually require an even wider dynamic range, such as 65-75 dB (1:1800 - 1:5600). When imaging such a scene with a 60 dB imager, either details in the darker areas get lost in the noise ("cut off"), or details in the brighter areas are lost in saturation, or both. This is shown in the figure 1.



Figure 1: Response of a CCD


Charge readout circuits, analog amplifiers, and A/D converters for real-time video typically limit the CCD signal further to a dynamic range of 8 bits (48 dB). This range can be extended to 10 bits, by employing appropriate analog processing and A/D converters; however, this is impractical for most applications. Another alternative type of circuitry employs non-linear transforms, such as logarithmic functions or "knee" curves, to compress the 1:1000 CCD output signal down to an 8 bit signal. While depicting most of the image information, such methods necessarily suppress details in the highlights.

The last limiting factor is the display (or frame grabber). The dynamic range displayable by normal CRT monitors, operating in a lighted room, is limited to about 1:100. An LCD screen is even more limited. The 1:200 or so signal which is generated by the video circuits is further reduced by the display. To optimize the display, the user often needs to adjust the contrast and brightness control of the monitor. A user who wants to display the image at its maximum contrast, usually sacrifices some of the dynamic range.

Adaptive Sensitivity™ Solution

i Sight has developed proprietary non-linear algorithms which reduce the dynamic range of the video signal without any loss of details. i Sight has implemented the algorithms in a 650,000 transistors full custom VLSI chip. The chip takes a wide dynamic range video input signal, and converts it to a 48 dB or less dynamic range signal, suitable for display on any CRT monitor without any loss of details. All i Sight cameras make use of this advanced video processor.

Ideally, the wide dynamic range input signal would be represented by 12-16 bits per pixel. However, as explained above, normal CCD chips cannot produce such wide dynamic range signals at video rates. Hence, i Sight takes advantage of multiple exposures. The same scene is imaged more than once (in most cases, two exposures are sufficient). One exposure is made at a low level of sensitivity (e.g. with an electronic shutter set at a short exposure time). That "short" exposure contains highlight details, but most very dark image areas are lost in the noise. A second exposure of the same image is taken at a relatively high level of sensitivity (e.g. at a long exposure time). The "long" exposure contains details of the darker parts of the image, but the brightest areas may come out saturated, without any details. Subsequent to the acquisition, the two images are combined in a special manner so as to produce a single, very wide dynamic range image. See figure 2.



Figure 2: Wide Dynamic Range response of a CCD

That wide range image is further processed, using i Sight's proprietary algorithm, to reduce the dynamic range down to a useful level - normally 40 dB, which can be displayed on a CRT monitor. A conceptual diagram of the i Sight camera is shown in the figure 3.



Figure 3: A conceptual diagram of i Sight camera

United States Patent 5,247,366 - Color wide dynamic range camera

Abstract:
The apparatus is a color wide dynamic range video camera which takes a plurality of images at different exposure levels, applies neighborhood processing to each of the images, and then combines the components into a final image.

Inventors:
Ginosar Ran; Zinaty Ofra; Sorek Noam; Genossar; Tamar ; Zeevi Yehoshua; Kligler Daniel J..

Assignee:
I-Sight Ltd.

Adaptive Sensitivity and the advanced digital video processor of i-sight’s cameras emulate the human eye's ability to perceive detail in high contrast lighting environments, and provide high detail video images even in the most extreme conditions. This unique capability offers the best video imaging system available for such applications as endoscopy, microscopy, machine vision, robotics, surveillance and many other areas of industrial process control and research.

Related Patents

Wide Dynamic Range Mosaic CCD Color Camera ( R. Ginosar, Y.Y. Zeevi D. Kligler, N. Sorek, T. Genossar and O. Zinaty), US Application, March 1993.

Wide Dynamic Range Camera (Y.Y. Zeevi, R. Ginosar and O. Hilsenrath), Israel, 1988, Worldwide 1989, U.S.A. Patent No. 5,144,442, 1992.

Wide Dynamic Range Mosaic CCD Color Camera ( R. Ginosar, Y.Y. Zeevi D. Kligler, N. Sorek, T. Genossar and O. Zinaty), US Application, March 1993.

Color Wide Dynamic Range Camera (Y.Y. Zeevi, R. Ginosar, D. Kligler, O. Zinaty and N. Sorek), U.S.A Patent No. 5,247,366, 1993.