[EMAIL PROTECTED] wrote: > I don't know what's causing that, but I've seen a bit of it as well. It > happened to me in a room with mixed fluorescent and tungsten lighting. Only > once have I seen it change from one shot to the next. I expect it will go > away with the first firmware update.
I've seen it with every digital camera I've used, and even see it happen to a lesser degree when using "P" or "Hyper-P" mode on film cameras. I suspect it's a combination of things. One of them is the "slice of a cycle" issue Godfrey described. However, I've seen it happen in a sequence of daylight shots taken outside in machine gun mode, so I don't think that's the whole answer, either. One thing I think plays a role is going to take me a minute to explain. Basically, AWB only has one thing to play with, and that's the actual content of the image as captured by the sensor and the metadata describing the parameters of the capture process itself. It runs some algorithms on this data and decides what it thinks the WB should be. I suspect that some of those algorithms involve evaulating the histogram. That means that changes to the content of the image as captured can change what it thinks ought to be the WB. As an example, look at the first three photos on the page you posted. http://users.accesscomm.ca/wrobb/temp/whitebalance/wb.html In the first one, it seems like it's picking a "more tungsten" WB. In the second one, it seems to be picking something closer to "flourescent" or "daylight". In the third one, it's somewhere in between. Now look in detail at the differences in what was captured in those images. The first one is a bit to the left of the next two, and doesn't include any of the door. The second one includes just the silver light switch and conduit and a thin slice of the dark door jamb. The third one contains most of the door, including that pink poster on it, and some of the white wall on the other side. My guess is that those differences in content are what make it pick different WBs in a situation like this. -- Thanks, DougF (KG4LMZ) -- PDML Pentax-Discuss Mail List [email protected] http://pdml.net/mailman/listinfo/pdml_pdml.net to UNSUBSCRIBE from the PDML, please visit the link directly above and follow the directions.
