Pentax Pixel Shift Resolution inside Affinity Photo? (posted to wrong thread yesterday)

[GMPhotography]

I posted this to the wrong thread yesterday, it was meant to be a question not a feature request.



I seem to have issues with Affinity photo and developed Pixel Shift resolution files from my Pentax K3II. I didn't realize what was going on right away, but after I diagnosed a different problem I was just having, I realized the error i was getting earlier was when i was editing a developed Pixel shift resolution image. Affinity would give me the following error, close to the same wording "There are no pixel elements in the project. Flatten and save anyway?"
Also, with my recent image, when I was done an exported, Affinty would add moire like patterns in the sky. I couldn't figure it out until i replaced the Pixel shift resolution foreground with the non-pixel shift resolution version and solved the problem.
Any ideas on this? Can Affinity not handle the type of image that is generated when I develop a PSR file from my Pentax K3II? It would be sweet if this could be solved in an update, and maybe add support for developing a PSR file from it's RAW version. The Pentax Digital Camera utility is clunky and annoying to use.

Here are the two different versions of that image. Go to Flickr to see full size:

Banded version: BuseAffinityFullEdit by Greg Murray, on Flickr

Clean version: BuseNonPSR by Greg Murray, on Flickr

[MJSfoto1956]

most RAW converters have trouble with Pentax (and Olympus) pixel-shift RAW. For example, DxO Optics Pro refuses to support these unique RAW formats (even though they *do* support pixel shift with their own DxO One camera -- go figure). I use DCU5 for my Pentax pixel shift images. Then these can be saved as TIFF. And then Affinity Photo can take it from there.

 

Michael

[GMPhotography]

most RAW converters have trouble with Pentax (and Olympus) pixel-shift RAW. For example, DxO Optics Pro refuses to support these unique RAW formats (even though they *do* support pixel shift with their own DxO One camera -- go figure). I use DCU5 for my Pentax pixel shift images. Then these can be saved as TIFF. And then Affinity Photo can take it from there.

 

Michael

This foreground was a tiff from DCU5. The image only became clean when I exported a tiff from LR (not DCU5) avoiding the development of the PSR. The sky was done with astrotracing therefore it is a regular image, not a PSR image. Mixing the two different types into one photo gave me major problems upon export, as you can tell.

[GMPhotography]

what I mean when I say I "Exported a tiff from LR": I developed the RAW PSR file and exported a tiff. When you develop a PSR file inside LR, it takes only one of the 4 images and uses that, so as if PSR wasn't even switched on at the time. This tiff worked great in affinity when combining it  with the tracked sky from that night. the DCU5 version, not so much.

 

This foreground was a tiff from DCU5. The image only became clean when I exported a tiff from LR (not DCU5) avoiding the development of the PSR. The sky was done with astrotracing therefore it is a regular image, not a PSR image. Mixing the two different types into one photo gave me major problems upon export, as you can tell.

[MJSfoto1956]

as I understand it, one cannot use Pixel Shift with Astrotracer (at least not on my K-3ii) -- so I must be missing something here.

 

Michael

[GMPhotography]

as I understand it, one cannot use Pixel Shift with Astrotracer (at least not on my K-3ii) -- so I must be missing something here.

 

Michael

yes, that is exactly what i said "The sky was done with astrotracing therefore it is a regular image, not a PSR image". the final image is a composite. If it weren't the landscape would be blurry and I wouldn't be using Affinity. as stated several times: I shot the sky with astro-tracer and the foreground with PSR. what is apparent is that you can't composite a normal tiff (sky) with a DCU5 developed PSR tiff (foreground) in affinity, there is clear conflict with the pixels upon exporting. Not really sure why, it would be good to know.

[anon1]

have you tried comparing pixel shift to stacking?

reviews I´ve seen so far always compared pixel shifted images to normal images which totally misses the point

 

it would be interesting to see how pixel shifted images stack up against two normal images stacked using median or average mode

- just saying cause I think the difference will be minor and you avoid the pixel shift problems

 

I´ve seen drastic improvements in stacked images for myself so this is definitely a great technique for everyone 

- not real higher resolution of course but the image is just more accurate and thus has more detail (more "true resolution" if you want to call it like that)

 

and while you´re at stacking you can just as well use 5 or ten images and I´d bet that a stack of 10 images has a higher quality/ more detail than a pixel shifted image out of 2 images 

[MJSfoto1956]

I've been doing layered Pixelshift/Astrotracer and BlendedStacked/Astrotracer images for a couple of years now. And I can assure you adding two TIFFs as layers in AP or PS (or any tool for that matter) will never produce the result I'm seeing in your post. But as I said before, many RAW converters have trouble with Pixelshift data and I assume the one built into AP likely does, so I only use those that produce good results (or as an alternative, I use a blended stack to achieve something similar). Given that my methodology starts with RAW conversion using the best RAW converters for the task at hand (e.g. DxO Optics Pro or DCU5 for pixel shift) I doubt I will be using Affinity Photo's built-in RAW converter anytime soon. I would suggest you process your pixel shift images separately using whatever tool gives you the best result, then import those TIFFs into AP as a layer and go from there. While a total "end to end" tool might be nice, I have yet to find anything remotely close and choose instead to use the best tool for the job. YMMV

 

Michael

[Leigh]

GMPhotography, I've just downloaded a Pixel Shift RAW sample and made similar edits and wasn't able to replicate the issue. Is it possible to send me the files you used via this dropbox link so I can look into this further?

 

Your files will remain private and I will make sure they're deleted when finished with :)

[GMPhotography]

GMPhotography, I've just downloaded a Pixel Shift RAW sample and made similar edits and wasn't able to replicate the issue. Is it possible to send me the files you used via this dropbox link so I can look into this further?

 

Your files will remain private and I will make sure they're deleted when finished with :)

Sure will Leigh

[GMPhotography]

I've been doing layered Pixelshift/Astrotracer and BlendedStacked/Astrotracer images for a couple of years now. And I can assure you adding two TIFFs as layers in AP or PS (or any tool for that matter) will never produce the result I'm seeing in your post. But as I said before, many RAW converters have trouble with Pixelshift data and I assume the one built into AP likely does, so I only use those that produce good results (or as an alternative, I use a blended stack to achieve something similar). Given that my methodology starts with RAW conversion using the best RAW converters for the task at hand (e.g. DxO Optics Pro or DCU5 for pixel shift) I doubt I will be using Affinity Photo's built-in RAW converter anytime soon. I would suggest you process your pixel shift images separately using whatever tool gives you the best result, then import those TIFFs into AP as a layer and go from there. While a total "end to end" tool might be nice, I have yet to find anything remotely close and choose instead to use the best tool for the job. YMMV

 

Michael

Perhaps you need to read more. I don't use the raw developer in AP for my Pixel shift, as I sadi several times. the pixel shift image I was using was developed to a TIFF in DCU5 and then I used that TIFF in AP which gave me the issues. Swap that PSR TIFF out with a normal image, and things are fine.

[GMPhotography]

GMPhotography, I've just downloaded a Pixel Shift RAW sample and made similar edits and wasn't able to replicate the issue. Is it possible to send me the files you used via this dropbox link so I can look into this further?

 

Your files will remain private and I will make sure they're deleted when finished with :)

Leigh, did you use DCU5 to develop the pixel shift RAW file? If you just used AP, it wouldn't develop and combine the whole image, it would just use one of the 4 images that the Pentax takes, thus turning it into a normal file.

I'll send you the RAW and the version I developed in Pentax's DCU5.

[MJSfoto1956]

Perhaps you need to read more. I don't use the raw developer in AP for my Pixel shift, as I sadi several times. the pixel shift image I was using was developed to a TIFF in DCU5 and then I used that TIFF in AP which gave me the issues. Swap that PSR TIFF out with a normal image, and things are fine.

 

Perhaps you don't understand the quote: "most RAW converters have trouble with Pentax (and Olympus) pixel-shift RAW". Needless to say, Lightroom would be included here. Why is Lightroom's inability to produce a decent TIFF from a Pentax a problem for Affinity????

 

Michael

 

 

[GMPhotography]

Perhaps you don't understand the quote: "most RAW converters have trouble with Pentax (and Olympus) pixel-shift RAW". Needless to say, Lightroom would be included here. Why is Lightroom's inability to produce a decent TIFF from a Pentax a problem for Affinity????

 

Michael

 

 

Wow. Just wow... That's all I have to say. Let's let Leigh take over now ok?

[GMPhotography]

have you tried comparing pixel shift to stacking?

reviews I´ve seen so far always compared pixel shifted images to normal images which totally misses the point

 

it would be interesting to see how pixel shifted images stack up against two normal images stacked using median or average mode

- just saying cause I think the difference will be minor and you avoid the pixel shift problems

 

I´ve seen drastic improvements in stacked images for myself so this is definitely a great technique for everyone 

- not real higher resolution of course but the image is just more accurate and thus has more detail (more "true resolution" if you want to call it like that)

 

and while you´re at stacking you can just as well use 5 or ten images and I´d bet that a stack of 10 images has a higher quality/ more detail than a pixel shifted image out of 2 images

 

Pixel shift resolution in the Pentax takes 4 "normal" images (as in each one stands on its own as a full image). It does this by moving the sensor by 1 pixel such that the entire colour data of the scene is captured at each. The goal of pentax's version is not so that you increase overall resolution, the K3II still puts out a 24 mega pixel image (unlike the Olympus) the goal is as I said, so the sensor captures the full colour data and there is no guessing going on. In normal cameras, the camera will take the neighbouring values of the pixels that surround say a red (so 2 green and 1 blue) and guesses as to what values should be there for what that red pixel records. This is the Bayer pattern that all cameras have. When the pixel shift resolution file from my K3II is developed inside pentax's DCU5, the image basically becomes debayerd and the full real colour data exits at each pixel. I have a feeling this debayering changes the pattern the pixels form so my feeling is that when trying to replace the sky in a Pentax pixel shift image with an image that is not pixel shift resolution (and thus the normal Bayer pattern of r g g b) there is a conflict when things are exported. Follow this article as it explains what the Pentax does and how it differs from say, Olympus, which is after a higher megapixel count than the sensor is natively capable of: http://www.pentaxforums.com/articles/photo-articles/how-pentax-k-3-ii-pixel-shifting-works.html

So simply put, stacking multiple images manually would not accomplish what the Pentax can actually do.

All this being said, since the camera records 4 normal frames, when you develop the pixel shift resolution file outside of DCU5, you get a normal image, as if PSR wasn't even switched on in the first place.

Don't listen to Michael, the guy has no clue how this works. Ive reported him and will be ignoring all of his replies, as everyone else should.

[anon1]

Pixel shift resolution in the Pentax takes 4 "normal" images (as in each one stands on its own as a full image). It does this by moving the sensor by 1 pixel such that the entire colour data of the scene is captured at each. The goal of pentax's version is not so that you increase overall resolution, the K3II still puts out a 24 mega pixel image (unlike the Olympus) the goal is as I said, so the sensor captures the full colour data and there is no guessing going on. In normal cameras, the camera will take the neighbouring values of the pixels that surround say a red (so 2 green and 1 blue) and guesses as to what values should be there for what that red pixel records. This is the Bayer pattern that all cameras have. When the pixel shift resolution file from my K3II is developed inside pentax's DCU5, the image basically becomes debayerd and the full real colour data exits at each pixel. I have a feeling this debayering changes the pattern the pixels form so my feeling is that when trying to replace the sky in a Pentax pixel shift image with an image that is not pixel shift resolution (and thus the normal Bayer pattern of r g g B) there is a conflict when things are exported. Follow this article as it explains what the Pentax does and how it differs from say, Olympus, which is after a higher megapixel count than the sensor is natively capable of: http://www.pentaxforums.com/articles/photo-articles/how-pentax-k-3-ii-pixel-shifting-works.html

So simply put, stacking multiple images manually would not accomplish what the Pentax can actually do.

All this being said, since the camera records 4 normal frames, when you develop the pixel shift resolution file outside of DCU5, you get a normal image, as if PSR wasn't even switched on in the first place.

Don't listen to Michael, the guy has no clue how this works. Ive reported him and will be ignoring all of his replies, as everyone else should.

 

yeah I know about democaising but a fact is that the "guessing" is actually done by very advanced algorithms and when you then stack 4 images it becomes very high quality (just no image is the same)

 

 

if you´ve got the chance to do a direct comparison between pixel shift and a stack of images I´d be very interested in seeing the result

[MJSfoto1956]

Um, GMP is quite wrong in saying I don't know how Pentax Pixelshift works. I'm a long-time Pentax shooter and I have even done the exact same kind of pixel shift + normal layering he is talking about (in fact doing long exposure night photography with astrotracer just like he has). And I've never experienced anything remotely like what he has encountered and am genuinely interested in getting to the bottom of it. I suspect user error, but perhaps there might be another answer. In spite of his thin skin I would in fact like to help. However, it is quite clear he has already decided that he is in the right and that I am in the wrong so there is nothing much else to say. I certainly have no interest in reporting him for his rude remarks back to me. 

 

Michael

http://www.jmichaelsullivan.com