[Oakland Rob]

I heard someone else notice this. Is this problem unique to K-1 PS images? Because I never heard of this issue on K-3 II pixel shift images.

I'd say blue is a bit brighter in some K-1s I tested, but within the range usually found in the variants of raw renderings. But some have posted elsewhere (unfortunately PS discussions are scattered in several subforums) examples of a much more dramatic shift. And I haven't had trouble with FastRawViewer or some other RAW converters either.

Haven't had any problem with K-3II PS images.

[Glorfindelrb]

Lr vs PDCU, pixelshift with MC on. No edit yet.

[fromunderthebridge]

I can confirm that pixel-shift does not work properly on PS CS 6 (13.0.1) and CR 7 (9.1.1.461). Based on my tests I'm not even sure the software handles more data than just the first embedded frame (I may be wrong though).

The 'right' workflow for pixel-shift seems to be currently DCU -> 16bit tiff -> PS. I can also confirm that based on my tests the subject has to be completely still (at least for close-ups, no surprise there), no wobble and camera on tripod. Motion correction does eliminate movement pretty well, but if subject moves you gain nothing. Also, for some edge cases the MC does not eliminate all pixel shift artifacts (not sure about the specific circumstances).

EDIT: by still I mean very difficult outdoors as the slightest wind throws off the pixel shift for flowers and such. Haven't tested with open landscapes, probably works better there.

[Oakland Rob]

I can confirm that pixel-shift does not work properly on PS CS 6 (13.0.1) and CR 7 (9.1.1.461). Based on my tests I'm not even sure the software handles more data than just the first embedded frame (I may be wrong though).

The 'right' workflow for pixel-shift seems to be currently DCU -> 16bit tiff -> PS. I can also confirm that based on my tests the subject has to be completely still (at least for close-ups, no surprise there), no wobble and camera on tripod. Motion correction does eliminate movement pretty well, but if subject moves you gain nothing. Also, for some edge cases the MC does not eliminate all pixel shift artifacts (not sure about the specific circumstances).

EDIT: by still I mean very difficult outdoors as the slightest wind throws off the pixel shift for flowers and such. Haven't tested with open landscapes, probably works better there.

I think you're incorrect.

If ACR didn't use all four images, you wouldn't see the blurs/artifacts you do see. If you use something to pull the four frames you'd see smaller light trails in each image as well; the length of the boat and the headlights are sort of like timers (notice how PDCU chopped out the curve in the headlights of the turning car for example).

Furthermore, you can't see much improvement with PS in ACR/Lr until you increase the sharpening significantly (and depending on what you do, obviously contrast, detail and other adjustments have to be used to discern the increase in resolution).

For those of us who like to use RAW, using PDCU isn't right. It can remove some movement, but it's a blunt instrument, and I for one find it atrocious to use. And I prefer raw and old fashioned retouching to fix the movement (although as you note, some movement like a little camera shake pretty much ruins the whole shot no matter what you use).

I'm awaiting a tool like some HDR programs that let you adjust ghosting control. You could remove variable amounts of movement. You may not need the higher resolution in lots of parts of your image. Many sample shots I see are landscapes, but if you're shooting a subject close and just want nice bokeh in the background zapping out the moving leaves there and leaving a lower resolution is fine.

[Glorfindelrb]

Ofc ACR use all 4 images and the sharpening is not the same there (by default) but there is still some issues. I'll try to expose them using comparing with MC off if I manage to find time for this.

Yes PDCU is a pain to use but removing every little artifact is a lot of work and not all pictures deserve that commitment (Ok maybe all 140Mo pictures but still). Having a good script doing the job and leaving a two layered file with the PS raw, the single shot extracted from the PS and a MC mask on it would be the best solution, it's not available yet.

[fromunderthebridge]

I think you're incorrect.

If ACR didn't use all four images, you wouldn't see the blurs/artifacts you do see. If you use something to pull the four frames you'd see smaller light trails in each image as well; the length of the boat and the headlights are sort of like timers (notice how PDCU chopped out the curve in the headlights of the turning car for example).

Furthermore, you can't see much improvement with PS in ACR/Lr until you increase the sharpening significantly (and depending on what you do, obviously contrast, detail and other adjustments have to be used to discern the increase in resolution).

For those of us who like to use RAW, using PDCU isn't right. It can remove some movement, but it's a blunt instrument, and I for one find it atrocious to use. And I prefer raw and old fashioned retouching to fix the movement (although as you note, some movement like a little camera shake pretty much ruins the whole shot no matter what you use).

I'm awaiting a tool like some HDR programs that let you adjust ghosting control. You could remove variable amounts of movement. You may not need the higher resolution in lots of parts of your image. Many sample shots I see are landscapes, but if you're shooting a subject close and just want nice bokeh in the background zapping out the moving leaves there and leaving a lower resolution is fine.

I am not seeing artifacts actually. I'm seeing what looks like a plain, old single frame. I don't know other issues different versions of LR/PS/CR have, but mine do not seem to react at all to PS. Although, I may be wrong as I have not done extensive testing... Just wanted to clear up that I'm not seeing artifacts, just seeing no effect at all. Will have to do the sharpening comparison though.

Btw. I also vastly prefer a non-destructive workflow and raws. Gives more room for me to play around and go back to 'old negatives' for tweaks.

[fromunderthebridge]

I am not seeing artifacts actually. I'm seeing what looks like a plain, old single frame. I don't know other issues different versions of LR/PS/CR have, but mine do not seem to react at all to PS. Although, I may be wrong as I have not done extensive testing... Just wanted to clear up that I'm not seeing artifacts, just seeing no effect at all. Will have to do the sharpening comparison though.

Btw. I also vastly prefer a non-destructive workflow and raws. Gives more room for me to play around and go back to 'old negatives' for tweaks.

My sincerest apologies, I was wrong, ACR actually does do something with the pixel shift data. *Commits ritual forum suicide* The difference is minute. Pixel peeping revealed increased detail in a macro shot of a fabric.

There are still rendering differences between DCU and ACR even when both are set to the most neutral settings possible. DCU seems marginally better wrt noise. If I disable noise reduction altogether in ACR I get color blotching, but not so with DCU. This is probably a separate issue and it may be that DCU does noise reduction regardless of the settings (ACR shows blotching with single shot and pixel shift at iso 6400).

[Oakland Rob]

Ofc ACR use all 4 images and the sharpening is not the same there (by default) but there is still some issues. I'll try to expose them using comparing with MC off if I manage to find time for this.

Yes PDCU is a pain to use but removing every little artifact is a lot of work and not all pictures deserve that commitment (Ok maybe all 140Mo pictures but still). Having a good script doing the job and leaving a two layered file with the PS raw, the single shot extracted from the PS and a MC mask on it would be the best solution, it's not available yet.

I thought if one bought a K-1 every image, like precious children, is special and worthy of as much attention as possible, as opposed to lowly APS-C non-PS riffraff

But I do agree. Using dcrawps with percentage thresholds for removing artifacts is pretty great, esp for really diffuse small stuff. But you have to experiment to get the right level, although once done it can work pretty efficiently. I think internally it essentially does what you're describing.