| Tonight's photo invention |
[Jul. 26th, 2008|12:15 am] |
On my drive home from work this evening, I was thinking about setting exposure in cameras. In short, camera light meters haven't really changed a lot in 50 years, despite the fact that your basic camera is ten million light meters. They are optimized for snapshots, or action shots, but pretty much suck for posed shots.
When I'm taking a picture that I care about, I'll fiddle with the light meter, take a guess (or let the camera take a guess) at the exposure, take a test shot, look for blinkies, check the histogram, readjust the exposure, lather, rinse, repeat. If I really care about the picture, I'll bracket. The general goal is to get the brightest point in the frame to be just under the point where it is totally over exposed (or clipping to use signal processing terms). In some cases, I may let some highlights clip so that I can get details in the shadows.
When I'm doing all this, I'm basically taking a SCWAG at the exposure, hence bracketing. My histogram is a combination of all of the colors, and if one color is oversaturated, it may not show up in the histogram, or the blinkies. The camera, on the other hand, especially if it had a CMOS sensor, could do this all, a lot more accurately.
Let's take that simple, non-HDR, case. The camera could take a picture at a nominal exposure. It now knows the exposure of every single pixel in the frame. It can calculate the value of the brightest pixel, and based on that value, set the exposure so that it is just shy of clipping, take a check shot, and then set that as the exposure, until it gets changed.
That's the most basic version, there are many possible enhancements. The first one would be to allow the photographer to set some percentage of pixels that can be clipped, either over or under. That way, if it's a dark room, and there is one small, very bright light, you can throw away detail on the light bulb, and get a decent exposure on the rest of the room.
The next level of enhancement would be auto HDR. The camera would go through the same routine as above, but do it to calculate the values just short of over exposure at one end, and under exposure at the other. When you click the shutter, it would take multiple exposures, at different shutter speeds, so that among the different shots, every pixel is properly exposed. You then use software in post processing to combine those shots and get one with a much wider dynamic range than the camera can get in one shot. People already do HDR multiple exposures, but the cameras don't automatically set them up in this way (so far as I know). There are, of course, issues, of something moves in the shot.
There are, of course, many more enhancements that could be made on these. For example, when shooting macro, have the camera automatically bracket either/both exposure and focus. Memory is cheap. Speaking of bracketing focus, I'd love to see something that uses a wide aperture, and multiple frames, scanning the focus, to develop a 3-D model of a scene, by post processing and determining what is in focus in which frame. |
|
|
![[info]](http://p-stat.livejournal.com/img/userinfo.gif){kind=small}
lrc's journal
