SD14 red layer versus SPP

richard stone wrote:

xpatUSA wrote:

Currently, my SD14 has the dust-cover removed and, for visible light shots, I put a Schott BG38 UV/IR blocking filter on the lens, went out and took a test shot. Because the BG38 sees a bit of near IR, I was looking at the red layer in RawDigger from the aspect of micro-contrast and focus. The focus was perfectly OK - but micro- or local-contrast was lacking, as is common in RawDigger layer views. When an image lacks micro- or local-contrast, it is often considered to be "soft" or even "out of focus"

Since most of the acutance we see in converted images is due to the converter processing, I though I would process the layer image best I could in RawTherapee and compare that with a default-processed SPP image. Here is the result:

Please view original size

The wires, the foreground tree and the transformer tell the tale, I reckon.

Also notice the overcast cloud detail revealed by RT and the great improvement of detail almost hidden by veiling flare.

In RT, the processing was mainly CBDL wavelet and RL deconvolution, with some impulse NR and Levels.

Ted

What is your conclusion, if any, after the work that you did?

I conclude that Foveon layer raw captures or Foveon raw composite images are lacking in acutance and are only acceptably acute after the quite drastic processing provided by SPP et alia.

How is what you did going to affect your decisions in terms of processing in SPP, and in RT (for example)?

My work flow remains unchanged for serious work: as neutral as possible in SPP, artistic/output adjustment in RT.

For me, after using SPP and RT on many sdQ files, it seems clear that SPP is probably the converter to use, aiming for a "neutral" file (as I recall you suggesting) to then send to RT, if the very best result is necessary or desired. But "neutral" would seem to mean reasonable levels of detail including some "noise," because if you end up at the most "smooth" setting in SPP or use "excessive" NR in SPP some (too much?) of the "detail" is lost along with the noise. And neutral would also seem to suggest not "over" sharpened. And "over," in this context, probably means less than the zero setting in SPP.

I don't bother much with noise in SPP (no choice with the SD14 under 400 ISO anyway).

Part of what we may be dealing with here is Sigma's need to find "default" and mid-range settings that a reviewer or first time user can use and get good (meaning decent or reasonable...) results right out of the box, so to speak.

Again, with the SD14, I doubt that Sigma is working on that.

I did a medium sized print of my Blue Boat and Beach image, after RT, in which you pointed out some noise in an out of focus grey-ish-brownish plastic jug at the back of the boat, and the jug looks like the right color and the "noise" in the enlarged screen image seems to have vanished. A triumph of the printer program, perhaps. It's a very sharp image.

Interesting, certainly some magic at work there, by the sound of it. The image of the jug that I posted was up-sized Nearest Neighbor so it does show the actual pixels rather than some smoothed or further-processed version thereof.

I appreciate the engineering and detail-oriented approach to this, but it seems plain to me that Sigma still has some work to do to get the most out of the Q-type sensors with SPP, if that is indeed the goal.