[stevebrot]

You would need to downsample the 24MP K-3 shot to 16 MP to match the K-5.

Yes, that is essentially true if you use that process. The full evaluation would involve a determination of captured detail between the two cameras at the same ISO and with the same output resolution. Noise is data lost.


Steve

[emalvick]

I only have a regular K5, but qualitatively I'm finding that the K3 (perhaps because of the resolution) is giving me the same amount of noise as my K-5 at say 6400 or 12800, but the extra resolution is giving me additional detail. For my happiness, the loss of detail is what really makes noise problematic. I don't care if it's there to some extent, I just don't want it eliminating important details. With my K-3, that isn't happening. Otherwise, noise is difficult to quantify, and I've certainly had shots over the years where my K-5 at ISO 400 was too noisy. Why? weather, shutter speed, light, etc... There are many factors besides sensors that will contribute to a noisy photo. I'm actually quite happy with my K-3 at this point.

I can see where under highly controlled conditions it might take a downsampling to match the K-5, but I don't need that.

[MJSfoto1956]

and lets not forget that most downsampling algorithms introduce a touch of sharpening, giving the downsampled image a further "boost" in apparent IQ vs. the non-downsampled image. IMHO, a better comparison is to UPSAMPLE both images to 20"x30" or even 24"x36" (common print sizes) then compare the two.

YMMV

Michael

[KDAFA]

and lets not forget that most downsampling algorithms introduce a touch of sharpening, giving the downsampled image a further "boost" in apparent IQ vs. the non-downsampled image. IMHO, a better comparison is to UPSAMPLE both images to 20"x30" or even 24"x36" (common print sizes) then compare the two.

YMMV

Michael

Interesting argument.

[aleonx3]

To be honest, I have not done enough yet to tell the difference between the two cameras in terms of noise level; but I do notice already the accuracy of WB on the k-3 (both set to Auto-WB on RAW) which I think is more important than noise level; sorry I digress...

[audiobomber]

and lets not forget that most downsampling algorithms introduce a touch of sharpening, giving the downsampled image a further "boost" in apparent IQ vs. the non-downsampled image. IMHO, a better comparison is to UPSAMPLE both images to 20"x30" or even 24"x36" (common print sizes) then compare the two.

Or downsample both to 8mp like DXOMark does for their "Print" view.

[emalvick]

and lets not forget that most downsampling algorithms introduce a touch of sharpening, giving the downsampled image a further "boost" in apparent IQ vs. the non-downsampled image. IMHO, a better comparison is to UPSAMPLE both images to 20"x30" or even 24"x36" (common print sizes) then compare the two.

YMMV

Michael

Good idea, but I can see a lot of problems there. It seems like upsampling a K5 image may actually enlarge the noise specks rather than increase random noise as the higher resolution sensor actually does.

The result is that these comparisons are probably far from perfect.

The ideal thing would probably be to make a print from each camera BUT make the prints sized differently at a similar ratio as their sensor resolutions. I think that's easier said than done (maybe a 4 x 6 vs 5 x 7 is close or really large like two sizes quote above). Then see how noisy the images seem.

---

Last edited by emalvick; 12-18-2014 at 03:44 PM.

[BrianR]

The ideal thing would probably be to make a print from each camera BUT make the prints sized differently at a similar ratio as their sensor sizes. I think that's easier said than done (maybe a 4 x 6 vs 5 x 7 is close or really large like two sizes quote above). Then see how noisy the images seem.

It wouldn't be hard, print the k-3 image at 8"x12" (or whatever) then print the k-5 image on an 8"x12" but with a thick border so the image prints in a 6.55"x9.83" area. This will be harder on the k-3 in the same way viewing on screen at 100% is since k-3 image is being magnified more. (I'm assuming here by 'sensor size' that you're referring to # of pixels and not a physical measurement.)

It makes more sense to me if you print them at the same size though. For the most part an incremental upgrade doesn't change end use. ymmv.

[emalvick]

It wouldn't be hard, print the k-3 image at 8"x12" (or whatever) then print the k-5 image on an 8"x12" but with a thick border so the image prints in a 6.55"x9.83" area. This will be harder on the k-3 in the same way viewing on screen at 100% is since k-3 image is being magnified more. (I'm assuming here by 'sensor size' that you're referring to # of pixels and not a physical measurement.)

It makes more sense to me if you print them at the same size though. For the most part an incremental upgrade doesn't change end use. ymmv.

I meant sensor resolution... I think many of us make that mistake... fixed above. If the sensor sizes were relative, we'd almost have our FF and the noise wouldn't really be an issue. That's really how Pentax ought to do it. These K3 and K5 sensors are so good on the noise that they could just stick with the pixel density and scale the sensor size up to FF.

[MJSfoto1956]

well, 20"x30" is my standard print size -- if a camera or lens cannot produce a good print at that size (regardless of its base resolution) then it is not something I will consider. In fact, this is the very test I use to determine if a body or lens is "good enough" -- I simply print it at 20"x30" and it either passes or it doesn't. I could care less about resolution, or "L glass" or "pixie dust" or whatever. I've printed my daughter's D700 images at 24"x36" (a mere 12Mp camera btw) and they look pin sharp. Likewise, on a lark I rented an Olympus OM5 last year and couldn't get a decent 20"x30" print that satisfied my requirements (in spite of all the glowing reviews out there). I'm a big fan of renting things before you buy -- over the years I have rented many different brands of cameras and lenses and frankly most have not been all that impressive to me. And I could care less what pixel peepers say. For me the gold standard is a large print.

YMMV

Michael