On Dec 14, 2010, at 7:04 PM, Matthew Hunt wrote: > On Tue, Dec 14, 2010 at 7:51 PM, Larry Colen <[email protected]> wrote: > >> Let's say that we're shooting through a crappy lens, or a dirty window, or >> we're taking a picture of a greycard on a beige sedan on a foggy day. There >> is plenty of light, but there is only 5 stops of dynamic range in the scene >> that we're shooting. I.e. all of the values fall between 2048 and 2080 mV >> at base ISO, or ETTR 4063 to 4095mV which get represented by the values >> 2048-2080 (or 4063-4095). > > You lost me here. If there's 5 stops of scene dynamic range, the low > and high pixel voltages should differ by a factor of 32; for example, > 64-2048 mV.
You're taking the five bits off the top, not the five bits off the bottom. But you've got a point there. > > I think what you're trying to do, specifically, is reduce quantization > noise--the noise caused by rounding to the nearest bit--by allocating > your whole 12-bit dynamic range only to whatever the scene dynamic > range is, without loss. My thinking is that that, especially if you > "expose to the right", quantization noise should be smaller than other > sources of noise, such as photon shot noise and amplifier read noise. > I don't, however, have figures for consumer sensors to back up this > intuition. That's pretty much my idea. > > -- > 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. -- Larry Colen [email protected] sent from i4est -- 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.
