On 01/09/06, Toralf Lund <[EMAIL PROTECTED]> wrote: > Actually, after thinking about this for a while longer, it occurred to > me that per-pixel gain and offset may be all there is to it. Such > adjustments can of course cancel out certain variations in the pixel > values that may be though of as "noise" (although I don't think it > really is noise from a signal-processing perspective), and thus also > improve the effective dynamic range somewhat.
Being able to sense and apply per-pixel gain will lead to a normalized output per pixel but it still can't negate thermal and other random noise sources. > I still think that using terms like "Dynamic Range Expansion" and > talking about noise-less pictures at high ISO is hyping up the > technology, though. By the same token, it seems unlikely that this would > make such a great difference that the 67 is rendered obsolete. (In this > context we must also take into account the fact that the setup may well > have had a worse noise performance than the 6MP cameras before this > circuit was added.) As you can likely appreciate 22bits is way over the top for most any analogue sampling system and in the case of these types of sensors probably at least 8 bits more than is necessary to capture the signal over the noise floor. But this I think is the key, the noise is sampled too and possibly due to the absolute integrity of the capture some usable information may be able to be extracted from the noise floor. The D range expansion really refers to non-linearizing the extremes of the capture range, IE softening the transfer curve at the noise floor and at saturation which should as suggested provide a more film like rendering. > And I don't understand why they call this an analog image processor, > either, since the block diagram indicates is done after an ADC stage... Remember the ADC is sampling discrete R-G-B samples from the sensor, post processing still has to occur to produce the final massaged RGB images (or encapsulated RAW data). > On the other hand, if this was what Aaron was talking about I think I'd > agree with him that it's a no-brainer feature. In fact, I've used > line-scan cameras that have such a setup earlier. I circuit like this > obviously costs something, though, so it's probably not that nobody have > thought about it earlier. It has to be a fast system so maybe it's just the first time a manufacturer has brought the technology into the realms affordability for inclusion in consumer items? > I'm assuming here that the DSLRs don't already do proper dark and light > calibration, though; I must admit I haven't really studied them hardware > close enough to be 100% sure. They do but I think it's a lot more rudimentary than the options that this chip offers. -- Rob Studdert HURSTVILLE AUSTRALIA Tel +61-2-9554-4110 UTC(GMT) +10 Hours [EMAIL PROTECTED] http://home.swiftdsl.com.au/~distudio//publications/ Pentax user since 1986, PDMLer since 1998 -- PDML Pentax-Discuss Mail List [email protected] http://pdml.net/mailman/listinfo/pdml_pdml.net
