NIT announces a range of color sensors operating in logarithmic response mode. These new sensors are based upon NIT patented Native WDR pixel technology where the pixel photodiodes are used as single solar cells providing extraordinary intrinsic dynamic range without exposure time.
The new sensors deliver sharp and accurate colour images over a dynamic range exceeding 140 dB, without any setup or need for white balance. Specific algorithms for colour de-bayerisation have been developed in order to accomodate the specific logarithmic response of NIT sensors. The photoelectric signal response of NIT photodiodes is predictable and modeled with extreme accuracy which simplify the chromatic calibration process and further increase the colour stability over temperature changes.
A dedicated Native WDR pixel of 5.6 um size, implementing micro lenses and a colour bayer array has been designed and implemented first on a D1 - 768x576 pixels sensor. Youtube video links are here and here:
Thanks to P.P. for sending me the info!
This is pretty cool. I'd like to see the color chart cropped and gained up to the same brightness in each picture to see how much noise is present when the light is on. I wonder if lens flare, not the sensor, becomes the limiter to SNR in the dark regions when the light is on?
ReplyDeleteThey must have some good flare subtraction algorithm to go with this sensor/lens.
Are these lights turned on to full power? If so, this is incredible.
There is no special algorithm for flare suppression, the video's show the raw image after de-bayerisation. The big light is turn on to full power to 300 W.
ReplyDeleteI may be very naive, but how does one measure system dynamic range? Using calibrated charts? or calibrated light sources? Or is there an optical way where you could generate intensity variations spanning a large range?
ReplyDeleteWe measure through more than 6 decades by using high power illuminators and an integrating sphere.
ReplyDeleteScott Campbell came up with a great way to measure DR accurately. Illuminate with a laser thru a pinhole making a Fraunhofer diffraction ring pattern. The amplitudes of the rings are well known mathematically. Just count the number of rings that are visible and it is pretty indisputable with a single image. This method was published in the IEEE Workshop on CCDs and Advanced Image Sensors in 2001 by Y. Wang et al. It should a standard image for all HDR/WDR presentations.
ReplyDeleteI really don't want to be a jerk but can someone point out to me how this differs from the demos we could do with three-tube vidicon color cameras in 1968? Is it just that such things are harder in silicon? I understand that these chips are smaller, more reliable, etc., but what is the actual functional improvement?
ReplyDeleteAlso, what causes the broad blue stripe below the halogen lamp?
Interesting question Dave. There probably are few people who know the specs of a vidicon tube and I am not one of them.
ReplyDeleteNevertheless, let's try the top ten:
1. More responsivity (volts/input watt)
2. Lower noise
3. Better SNR at a given scene light level
4. No lag
5. Increased dynamic range
6. Less power
7. Higher resolution
8. Camera is smaller than a breadbox
9. Smaller pixels at same resolution means smaller lenses so weighs less
10. Lower dark current (I think)
The interesting part of the question is the ranking in importance. I did not attempt to put these in order.
Hi Dave,
ReplyDeleteWhat is the experience with Tri-vidicon tubes camera please ?
-yang ni
The sensitivity in vidicon tubes (As2S3 photoconductive target material) is controllable by changing the voltage across the target material. The range of this control is such that a vidicon can image in normal room lighting and in sunlight with no other adjustments. Automatic control requires only detecting the signal amplitude somewhere in the chain and using this to control the target voltage.
ReplyDeleteIn addition, the photoconductive layers have a gamma of about 2/3 so they inherently compress, say, 6 orders of magnitude of illumination to 4 orders. It was quite common with such cameras to be able to see out a window and inside the room at the same time. This looked very much like your image sequence with the halogen lamp on. Moving the window out of the field of view raised the target voltage and then the sene looked like the your image sequence with the lamp off.
In fact, we used to try to use early CCDs for license plate reading and had unending problems with saturation whle a vidicon camera could view the scene without trouble.
Of course, CCDs (or CMOS imagers) do have all of the benefits that Eric listed but I am just surprised that anyone thinks that achieving this dynamic range is in itself any big deal. Maybe it is just because I am an old guy.
Thanks Dave for kind reply !
ReplyDeleteI agree with you that "old" imaging tubes have a lot of intrinic advantages that solid state imagers cannot have. You can also mention:
1. SuperOrthicon which gives a logarithmic response
2. SEC imaging tube which has virtually no dark current and can integrate over a long duration
etc ...
What we would like to show here is a new imaging concept which is, to our knowledge, not yet explored yet. This imaging technology gives an instant dynamic range more than 120dB with a constant contrast sensitivity.
The highly predictivity of the photoelectric response over this dynamic range is, I think, new and can be useful for a lot of applications.
You are right, a pure dynamic range is meaningless, many other properties in an imaging device should be taken into consideration.
-yang ni