h a l f b a k e r yIt's the thought that counts.
add, search, annotate, link, view, overview, recent, by name, best, random
news, help, about, links, report a problem
browse anonymously,
or get an account
and write.
register,
|
|
|
On many digital cameras, the total latitude within a picture is about six-seven f-stops. Anything that's too light will get blanked out completely, and anything that's too dark will get lost in the "mud".
My proposal would be to have an option to use different exposure settings for the "even" and
"odd" rows of pixels on the imaging element. For shots in favorable conditions, both settings would be identical. But when contrast would otherwise be overpowering, the even rows could be set 1, 2, or 3 f-stops lighter than the odd rows. If the camera's usable range was otherwise six f-stops but the even rows were pushed by two f-stops, there would then be a two-f-stop range where only the even rows were usable, a four-f-stop range where all rows were usable, and a two-f-stop range where only the odd rows were usable.
The net effect would be that the areas of extreme lightness and darkness would only have half the resolution of other areas, but they'd at least show up. Given that many cameras these days have sensor resolution that exceeds the usable quality of their optics, the resolution loss should not be particularly objectionable--especially if it avoids having parts of the picture get lost entirely.
Fuji SuperCCD
http://www.dpreview...2fujisuperccdsr.asp
"Beneath each microlens on the sensor surface (a photosite) are two photodiodes, the primary captures a dark and normal light levels (more sensitive), the secondary captures brighter details (less sensitive)." Not dynamically adjustable as per supercats idea, but a higher dynamic range, nonetheless. [st3f, Jul 06 2006]
Please log in.
If you're not logged in,
you can see what this page
looks like, but you will
not be able to add anything.
Destination URL.
E.g., http://www.coffee.com/
Description (displayed with the short name and URL.)
|
| |
I'm not sure how you intend to apply a different f-stop to alternating lines of pixels - since the whole chip is behind the same aperture.
To get the picture, you would require either: |
|
| |
(a) Two exposures of the same image (Such as when bracketing to test exposures) and some digital image processing, or |
|
| |
(b) Taking the meter reading for the bright background and then filling in the front with flash. |
|
| |
What might work in rare days of bright sunshine, would be a modification to the Bayer matrix: instead of two greens, you have a denser green and a normal green - the denser one acts as if an ND filter were present. This would give a luminance difference, which might be useful for extrapolating tonal difference. Maybe. Of course, this would lower the overall sensitivity of the matrix, which would mean that on normal gloomy days the camera is even more useless unless someone invents a pocketable full-size full-weight tripod. |
|
| |
The exposure time would be different. While you are correct that the term "f-stop" is generally used to describe aperture, the term is also used in some contexts to describe a 2:1 difference in exposure whether that results from a difference in aperture, exposure time, or neutral-density filtering. |
|
| |
Hmm... I'm not sure I'd go for using two greens in the Bayer matrix; a better idea might be to replace one of the greens with a pan-chromatic luminance pixel. Such a pixel would be inherently more light-sensitive than a color-filtered one, making it inherently useful for capturing luminance detail in shadowed areas. |
|
| |
Actually, though I glossed over the color-matrix issues, dividing the screen into even and odd rows would require replacing the Bayer matrix with some other color arrangement. Using rows of a repeating RGB pattern with the pixel spacing being 1.5 times as dense as the row spacing, and with alternate rows staggered by half a pixel, would probably work well. |
|
| |
Note that dividing the pixels into even and odd rows probably isn't the best way of divvying them up from a photographic standpoint, but from a chip-layout standpoint it seems much more workable than trying to do an interleaved checkerboard or other such thing. |
|
| |
Does nobody else have difficulties with the limited exposure latitude of most cameras? |
|
| |
Use monochrome for the dark areas? Or is that what you already said? |
|
| |
I (and no doubt, so would you) predict that the next technological molehill and resulting consumer onslaught, once the fuss about megapixels is seen for what it is*, likely to be which manufacturers sensor and firmware gives you the deepest dynamic range. Developments will likely be made in synthesizing a 'soft knee' response to overexposure or softening highlight clipping, and stances will be made in terms of how much wider the latitude of future sensor products are than their competitors, or than that of film. Of course, that's a difficult thing for the average general public to comprehend in a single take, whereas 'my numbers are bigger than yours' are always easy to grasp and are what separates real men from those who know their stuff and just get on with it. |
|
| |
* ie, anything between 2 and 12 is indistinguishable for most people's needs, most of the time, considering what most of them do with most of their snapshots, and sales hype over higher megapixels is little more than willy-waving. |
|
| |
I'm not in favour of the idea of being
able to set the exposure differently on
different parts of the image sensor. It
sounds too complex and... erm...
expensive. I am in favour of some pixels
being more sensitive than others to give
a better exposure latitude. |
|
| |
That said, most digital cameras crank
up the contrast and give you a picture
that doesn't reflect the full exposure
latitude of the sensor. (still sometimes
too narrow and, as Ian points out, with
much harder cut-off points than film). |
|
| |
Actually, thinking about it, didn't Fuji do
this with a strange octagonal grid that
incorporated RGB and low-light
sensors? (scurrys off to find link). |
|
| |
st3f: What would be overly confusing about having a setting "Dynamic range enhancer: Off, Low, Medium, High, Max" to specify whether the system should have a 7 f-stop range at full resolution, an 8 f-stop range with the middle 6 being at full resolution, or 9/5, 10/4, or 12/2. Doesn't seem that hard to me. |
|
| |
Basically, me idea would be much like the SuperCCD, but with the ability to tailor the response according to requirements. |
|
| |
I meant complex in terms of
manufacture.
Flicking the sensitivity of rows or cells
on a
chip sounds very difficult. Making a
sensor with a fixed but higher dynamic
range and dealing with that data in or
out of camera an awful lot easier. |
|
| |
There is a quality trade with this
approach, though. If you
are
using the same number of bits to
represent a larger dynamic range, you
run the risk of seeing posterisation (I'm
sure there's a better word for that) in
smooth parts of the image. |
|
| |
I don't know how the exposure-control signal is normally distributed on a CCD, but I wouldn't expect having alternate rows driven from different enable signals to be particularly difficult. |
|
| |
Hmmm... maybe not. [flings croissant at
supercat.] |
|
| |