h a l f b a k e r ySugar and spice and unfettered insensibility.
add, search, annotate, link, view, overview, recent, by name, random
news, help, about, links, report a problem
browse anonymously,
or get an account
and write.
register,
|
|
|
Most video cameras these days can accommodate a very wide total dynamic range from candle-light to full daylight. Unfortunately, they cannot very well shoot scenes which are in parts very bright and in other parts very dim. Either the bright parts come out supersaturated white or the dim parts come
out black.
What I would like to see is a camera which could shoot 90 frames/second in groups of 3, where in each group one picture would have an effective shutter speed of 1/200 sec, the next would have an effective shutter speed of 1/1000 sec, and the last would have an effective shutter speed of 1/40 sec. The three pictures would then be blended electronically, using the 1/1000 sec frame for the parts of the picture that were brighter and the 1/40 sec frame for the parts that were dimmer; output would be a single 30 frame/second YUV or NTSC video signal.
Arbitrary focus from multiple images
http://www.spie.org...i_jan98.html#VCP-01 If you can get arbitrary focus, I would assume you can get arbitrary contrast in a similar manner. This system uses a beam splitter and three cameras, because you can't change the optical focus at video rates. [rmutt, Dec 28 2000, last modified Oct 04 2004]
[link]
|
|
...Supercat is proposing a camera (singular), and not cameras (plural)... |
|
|
PeterSealy needs to write "I will read the idea before calling 'Baked!' and describing some tangentially related but entirely different he heard about on the Discovery channel once" a few dozen times on the blackboard. |
|
|
No beam splitter is needed, the "underlying problem" of perspective does not exist, and there is no "shutter assembly/film unit" because we're talking about a *video* camera (this is in the title of the idea, for crying out loud!), which means that there is neither film nor shutter. (The "shutter" on a video camera is generally an entirely electronic arrangement.) |
|
|
Supercat is proposing that three images be taken in *series* by a video camera. 1/200 + 1/1000 + 1/40 ~= 1/40, so it wouldn't even reduce the frame rate. As far as I know this would work and isn't baked; but there may be some reason having to do with CCD behavior and the concept of "shutter speed" that makes it difficult. |
|
|
In general, what Supercat is trying to do is increase the *dynamic range* of a camera, which is quite laudable; many video cameras can record admirably in low light conditions, but their dynamic range is unfortunately still quite meager. |
|
|
Yes, but so is the dynamic range of the video signal, as well as any phosphor/LCD system upon which you might be able to display it. To get better dynamic range, we need full-digital capture with at least 24 bits/pixel (more than the human eye can see) as well as something that can reproduce it, perhaps like the DLP (micromirror) technology. |
|
|
egnor: You understand and describe perfectly the notion I'm describing. A CCD would need to be modified to use this technique, though from what I understand of how CCD's work the modification would not be overly difficult. |
|
|
rmutt: The contrast would have to be compressed to be accommodated on a standard video signal; most likely someone would then have to adjust the contrast on the recorded video tape to achieve something aesthetically pleasing (the recorded signal would consist mainly of tonal values ranging from sorta darkish to sorta lightish; the raw tape would look washed out, but everything would at least be visible. The person viewing the tape would then be able to adjust the gamma curve to yield nice looking results). |
|
|
petersealy: The real-time processing hardware would be significant, and even 5-6 years ago would have been prohibitively expensive. Today, however, it wouldn't be too bad. BTW, my first idea on this concept was to use a beam splitter (and for shooting 35mm it might be interesting to have a camera that used a beam splitter to record two perfectly-registered frames one of which got 1/5 as much light as the other). Note that near-perfect registration would be needed for this trick to work; using multiple cameras would not be an acceptable option. |
|
|
To increase the dynamic range of a digital-based camera (video or otherwise): I haven't studied digital cameras, but I'd imagine that there's an array of elements on the back of the camera that each sample the light hitting them. If one built that array of elements with mixed responces to light (perhaps 30% best sensing dim. 30% best sensing medium and 30% best sensing bright) and all the numbers were saved, an optimum-exposure type picture could later be easally generated by taking the numbers in a region that weren't maxed out, and putting them in their correct range (according to the type of sensor). I know this is (at least biologically) baked within each and every eye. |
|
|
Firstoff, three sperate cameras would be completely impossible. No matter the variations in their shutter settings -- the lenses would not be viewing the scene from the same point, so unless you want a nifty spin-round or some sort of 3d effect, it wouldn't be much help. |
|
|
This idea seems feasible, but likely that it would be more than a little too expensive for the consumer range of video equipment. (Even if not too expensive to install, the companies will add a huge markup -- something like this could increase the quality of video dramatically.) |
|
|
I do a lot of television production, and will ask a few of my camera-obsessed buddies for their views on this subject. |
|
|
//I haven't studied digital cameras, but I'd imagine that there's an array of elements on the back of the camera that each sample the light hitting them.// |
|
|
On a CCD, as with a film camera, it's possible to adjust exposure time to control sensitivity. While the CCD "shutter" is electronic (essentially all of the light-sensing elements are kept drained until it's time to "open" the shutter, then shortly after the CCD stops draining them it transfers the picture data to a set of analog shift elements which then output the picture data sequentially. |
|
|
Since clocking out a full CCD worth of picture data in 1/1000 second would be impossible, some tricks would be required to make things work effectively. I don't want to go into the details of how I'd do this, but the amount of additional circuitry required in the CCD itself would be modest. The biggest additional circuitry expense would be the frame buffers to merge the three images, but with today's technology the price for that would be fairly modest. |
|
|
I would expect that incorporating all the circuitry required to do this would add about $50-$400 to the cost of a camera (price depending upon quality of implementation, etc.) and thus increase MSRP by $150-$1500. Something feasible on the very high-end consumer models, and quite reasonable on professional models (which can run $5,000-$10,000 or so). |
|
|
Video camera dynamic range is improving markedly with each passing year. Hasn't caught up to film yet, but it is getting close. Star Wars Episode II has been completely lensed with a Sony HD camera--no film at all. |
|
| |