The A7R3 and the Result of Pixel-Shifting

[Michael Erlewine]

Here is the article on Pixel-Shifting:

https://petapixel.com/2017/11/18/testing-sonys-new-pixel-shift-feature-a7r-iii/

Ony's New Imagine Edge Software can now be downloaded here:

https://support.d-imaging.sony.co.jp/app/imagingedge/en/download/


They have posted what I imagine is the final software for processing the pixel-shift images. The pixel-shift technique in the camera produces four raw images. Are we supposed to post-process those images before we submit them to the software?

My main question is this: If the cameras takes four images in raw and then processes them, what format are they returned in: JPG or TIF? I guess it can't be raw, but that is what I would like to have, of course. Anyone know?

[Michael Erlewine]

Someone on another forum sent me the answer. There was a side-link I missed.

https://support.d-imaging.sony.co.jp/app/imagingedge/en/instruction/1_2_about_psm.php


If I read that right, the final conglomerate images is raw, which is what I hoped it would be.

[Paul2660]

From reading further on this process, it appears that the Sony software that combines the files, (4 shots taken) outputs as a raw file.

Thus the process that Pentax did with the K1 in camera is done by the software on a PC with the Sony.

Question is still whether the Adobe support for this file will be any better than what Adobe did for the K1.  Adobe's support for the K1 pixel shift was the worst of all solutions, and was never improved.  Silky Pix, Iridient, and rawtherapee all did a better job than Adobe, but do not offer the DAM or tool set Adobe has IMO. 

Sony may very well have the impetus to get Adobe to spend more than 5 minutes on the conversion process and actually produce a quality file.

Pentax obviously did not.

Paul Caldwell

[Michael Erlewine]

From reading further on this process, it appears that the Sony software that combines the files, (4 shots taken) outputs as a raw file.

Thus the process that Pentax did with the K1 in camera is done by the software on a PC with the Sony.

Question is still whether the Adobe support for this file will be any better than what Adobe did for the K1.  Adobe's support for the K1 pixel shift was the worst of all solutions, and was never improved.  Silky Pix, Iridient, and rawtherapee all did a better job than Adobe, but do not offer the DAM or tool set Adobe has IMO. 

Sony may very well have the impetus to get Adobe to spend more than 5 minutes on the conversion process and actually produce a quality file.

Pentax obviously did not.

Paul Caldwell

My view is that the post-processing was important mostly for resolving motion-errors. In my case, as a still-life photographer, that is not an issue, so I can't image they could screw this up, but I maybe overconfident. As a focus-stacker, all of this post work becomes more difficult on top of processing a stack, etc.

[Paul2660]

Good point, for still life this should be just like the K1, which works great with LR/ACR in pixel shift mode as long as there is no motion, thus no aliasing.   

I guess my main point is off topic, in that the K1 showed over 1 1/2 years ago, just how good the solution of shifting (pixel shift) was:

Great detail, lower noise, better overall clarity,

But Pentax never got the mainline raw converters LR or C1 to get a good conversion (in LR's case) or any conversion in C1's case.  The fact that lesser cost products which are mainly supported by 1 person, rawtherapee and Iridient both could do it and superior to LR's one and done attempt, is unfortunate. 

Now that Sony has entered the game, things may be different. 

Paul Caldwell

[shadowblade]

Does anyone know if the four exposures (or at least the first one in a set) are also saved as individual RAWs, able to be processes individually instead of only in pixel-shift mode? That would make pixel shift useful in a much greater range of situations - use the pixel-shift image where there is no movement and blend in the single image, processed identically, to cover areas affected by motion blur (which, in most applicable scenarios,  will be small areas, e.g. moving leaves on a tree).

The two real game-changers would be:

1. If the software (current or future) could automatically generate a masking layer to detect movement, by detecting pixels showing discrepancy between the different exposures. This could then be used to blend a single exposure into the affected areas, takimg advantage of pixel shift where there is no movement, while avoiding motion blur artifacts (it could also be done manually by combimg through the image to find affected areas, but an automated process would be much faster).

2. If Capture One picked up on this amd implemented it in their software. Much better to do this in a powerful RAW converter than to have to do it in a less-powerful tool lacking many of the features and controls you'd otherwise use during conversion.

[Michael Erlewine]

The Sony A7R III arrived today. The menu system is the usual nightmare. What I need is a list of how to set it up instead of figuring it out all over again. So far, I can’t determine yet how to magnify the view so I can shoot from it. It’s not in the same league as the new Nikon D850 except perhaps in the pixel-shifted image, which of course, is why I bought it.

If I had any sense, I would send the whole thing back and just use the Nikon D850. LOL.

The pixel-shift on the A7R3 is easy to use, but for a focus-stacker very tedious, physically, but they seem to have done it right. The results are outstanding. Scary. Part of me would like to just send it back, but I probably won’t. In other words, it is not a lot of fun so far, especially since Adobe has not yet supported it. The thought of doing a 100-layer stacked image is daunting even to consider. Ouch!

The tiny paper manual ONLY tells where to look at an online version and when I do, there is about nothing there that really explains much of anything. We are really on our own. You would think that with all that money Sony must have, they could hire someone like Thom Hogan, etc. to write it right. Someone will make a fortune making it easy to know how to use this camera.

Looking at a pixel-shift image (my main interest) in Sony Imagining Edge Software is upsetting. For example, comparing the color in an image (pixel-shift) and the same image exported to PS as a TIF is troubling. The color is enough different between the two to make me wonder. The enclosed image (just a screen grab) shows the TIF colors in PS are more balanced than the same image at the same 100% in the Imaging Edge Software. The TIF is much better-balanced color. I must be the Sony Software as opposed to the image itself.

I guess I’m getting old, not only physically, but tired of endlessly moving to new equipment with all the attendant difficulties. I just want to take photos, but I can see this is a never ending stairway and perhaps not going to heaven, either. LOL.

With all my griping, the pixel-shift image at first blush is damn good. I have no excuse to send it back! Not being much of a field-shooter anymore, I don’t like the dinky-ness of the A7R3.

Anyway, I have the A7R3, am (sort of) getting used to it and am afraid I will have to keep it. LOL. We will see.

[Kevin Raber]

I did an article on Pixel Shift HERE.  I am not sure you are going see the Pixel Shift feature as part of C1 or LR.  As far as menus and set up I am scheduled to do a video on that next Monday and will post it soon after.  I have had my a7r III for a while now, plus was shooting with it prior to delivery.  It's an incredible camera.  I have set up menus and buttons for several functions plus customized the FN Button and choices.  May I suggest you download the manual as a PDF from Sony support.  I have mine on my iPad and it easier to read and bookmark.  Plus depending on what reader you use you can make notes in the manual with the Apple pencil.

My article clearly shows results of movements during the exposure like grass or leaves. The image quality changes may be hard for most people to see unless you go to making a  print.

Also, the software is not the most intuitive.  I use Capture One with the a7r II and a9.  Results and speed are super.  C1 v 11 is now available.  My desk is loaded with Sony cameras right now.  The RX0 and the RX10 iV.  So, I'm bust doing a lot of shooting.  I will be doing more Pixel Shift images to share.

[shadowblade]

I did an article on Pixel Shift HERE.  I am not sure you are going see the Pixel Shift feature as part of C1 or LR.  As far as menus and set up I am scheduled to do a video on that next Monday and will post it soon after.  I have had my a7r III for a while now, plus was shooting with it prior to delivery.  It's an incredible camera.  I have set up menus and buttons for several functions plus customized the FN Button and choices.  May I suggest you download the manual as a PDF from Sony support.  I have mine on my iPad and it easier to read and bookmark.  Plus depending on what reader you use you can make notes in the manual with the Apple pencil.

My article clearly shows results of movements during the exposure like grass or leaves. The image quality changes may be hard for most people to see unless you go to making a  print.

Also, the software is not the most intuitive.  I use Capture One with the a7r II and a9.  Results and speed are super.  C1 v 11 is now available.  My desk is loaded with Sony cameras right now.  The RX0 and the RX10 iV.  So, I'm bust doing a lot of shooting.  I will be doing more Pixel Shift images to share.

Are the individual exposures that make up the pixel shift image still available to be processed individually, as a non-pixel-shift image? Or are they permanently locked into the pixel shift structure, only able to be used as part of a pixel-shift image?

The former would be a lot more versatile - use the pixel-shift image for most of the scene, then paint over the motion-blur-affected areas with a single exposure to hide the artifacts - and would make more sense, given that the file isn't generated in-camera, but only in post-processing software. But I can't find any confirmation one way or the other.

[Michael Erlewine]

I did an article on Pixel Shift HERE.  I am not sure you are going see the Pixel Shift feature as part of C1 or LR.  As far as menus and set up I am scheduled to do a video on that next Monday and will post it soon after.  I have had my a7r III for a while now, plus was shooting with it prior to delivery.  It's an incredible camera.  I have set up menus and buttons for several functions plus customized the FN Button and choices.  May I suggest you download the manual as a PDF from Sony support.  I have mine on my iPad and it easier to read and bookmark.  Plus depending on what reader you use you can make notes in the manual with the Apple pencil.

My article clearly shows results of movements during the exposure like grass or leaves. The image quality changes may be hard for most people to see unless you go to making a  print.

Also, the software is not the most intuitive.  I use Capture One with the a7r II and a9.  Results and speed are super.  C1 v 11 is now available.  My desk is loaded with Sony cameras right now.  The RX0 and the RX10 iV.  So, I'm bust doing a lot of shooting.  I will be doing more Pixel Shift images to share.

Thanks. I had previously read your article and learned from it. I have no trouble with understanding what pixel-shift is, since I used it extensively in the Pentax K3 and K1 cameras. Of course, it is frustrating that Adobe has not yet supported it, but I have the Imaging Edge software and the Sony free version of C1.

My problem, as primarily a focus-stacker, is how to use this technique in the A7R3 as part of my work process.

Did I understand what little there is from Sony about this technique that Imaging Edge will take one pixel-shift photo, figure out where all four of them are, and process that stack to deliver to us the combined photo? That would help. Then I have to turn each combined photo into a TIF file to send it to Zerene Stacker, my stacking software of choice. As you (I'm sure) already well understand, processes like these extend themselves until that is a lot of what we do. If I process a 100-layer stack, this a long process.

I do have a question, not about pixel-shifting, but how do I with Imagining Edge and C1.... batch process a group of A7R3 images, convert them to TIFF, and save them to a folder. I am on a PC. I have several Macs, but don't use them for this kind of work. PC has always been faster IMO, much less expensive, and I can tinker with them myself.

I am uncharacteristically grumpy about the menu system of the A7R3, although I had two copies of the A7R2, etc. Probably because in the last couple of years I have the Pentax K3 and K1, the Hasselblad X1D and the Fujifilm GFX, all of which I had to learn and work with. I am kind of tired of new cameras, but can't help myself but check them out. IMO, and this is just me, the A7R3 is like a toy (a good toy) compared to the Nikon D850, which I love. 

Shadowblade: the four pixel-shift images are raw and available like any other A7R3 images.

[shadowblade]

Shadowblade: the four pixel-shift images are raw and available like any other A7R3 images.

Great - pixel shift just became a whole lot more useful. No longer just for still life and product shots in studio.

[Michael Erlewine]

Here is a proof-of-concept stacked image taken with the Sony A7R3 using pixel-shift as follows:

This is what I call a “short stack,” in this case six pixel-shifted images (each image using four pixel-shifts). These six resulting TIF images were then stacked using Zerene Stacker to produce a final stacked image.

Because I only took six images, there are areas in the photo that are not in focus, perhaps most easily seen in the Spadix (large vertical in center), where I focused on the bottom and top of the Spadix, but not in the center. If I were to take a more normal stacked image, it might have 50 or more layers, but here, to save time, I just made six images.

The result shows me that the A7R3 will work well for focus stacking, especially if they ever automate it better than they have now. I had to do each of the six images separately, since I don’t know how (or if) there is a way to batch process these at this point. I am impressed by the color and (unlike my Nikon D850), the A7R3 did not require much of any color adjustment. Interesting.

I was using the APO-El Nikkor 105mm on the Cambo Actus, with the Sony E-mount, and a little time delay. So, to repeat myself, I see no reason why this setup will not be excellent for focus stacking. However, to be really useful, we would need software that accessed a folder and automatically processed all the 4-image pixel-shift packets in that folder and (for me) then converted the composite images (in this case six) into TIF files, ready to stack in Zerene Stacker.

Aside from adjusting the exposure level, I did not post-process this image much at all.

Sorry if this is a little sloppy, but I am doing it for me and sharing it with you, if you can find it helpful.

[Michael Erlewine]

Here is an single image (unstacked) at f/16 to give you an idea of what the A7R3 can do. This is pixel-shifted using the Voigtlander 65mm Macro lens for Sony E-mount.

[Paul Roark]

>... the four pixel-shift images are raw and available like any other A7R3 images.

Great - pixel shift just became a whole lot more useful. No longer just for still life and product shots in studio.

Could you elaborate on this thought?

Thanks,

Paul
www.PaulRoark.com

[shadowblade]

Could you elaborate on this thought?

Thanks,

Paul
www.PaulRoark.com

Pixel shift is great for things that don't move, but creates terrible artifacts where there's any movement - clouds, leaves, grass, waves, etc.

In any given landscape, most elements don't move - tree trunks, rocks, etc. Therefore, by being able to isolate a single exposure from the pixel shift set and developing it separately, you can take advantage of pixel shift to increase detail in nonmoving areas, while masking out moving areas and painting over them with the single exposure. So you end up with pixel shift detail in the nonmoving areas, eliminating the motion artifacts while still retaining single-exposure detail in those areas.

[Michael Erlewine]

I am “trying” to test out the Sony A7R3 in single-shot and pixel-shift mode, testing it against itself and also as close as I can to see how it compares to the Nikon D850.

These shots are taken on the Cambo Actus Mini technical camera, using the D850 and A7R3 cameras in as close to the same conditions I can manage.

They were shot within minutes of each other in natural light coming in a window on a snowy and overcast day. The lens is the APO-El Nikkor 105mm lens at f/5.6.

The larger image is just an overview shot, a pixel-shifted image from the Sony A7R3 to provide context.

There are two crops:

One is with the Nikon D850 “as taken” and a second shot as cleaned up in post. Obviously, the second image could be to whatever taste we want. This comparison does show me that the Nikon is similar (to a degree and IMO) to what we call S-log in video. It is a bit of a canvas that we must tweak and finish in post, while I notice that the images from the A7R3 are much more finished and ready-to-go. I may prefer the S-log qualities; not sure yet.

And the second image compares a single image from the A7R3 with an image that has been pixel-shifted. As for this, I don’t see all THAT much difference between the two for all the pixel-shifting work. So I hope others can test this themselves and weigh in here with your thoughts. Yes, I see some differences, but not as much as I expected.

Of course, I have a lot more to do to form my own opinion of what I am seeing.

[Paul Roark]

Pixel shift is great for things that don't move, but creates terrible artifacts where there's any movement - clouds, leaves, grass, waves, etc.

In any given landscape, most elements don't move - tree trunks, rocks, etc. Therefore, by being able to isolate a single exposure from the pixel shift set and developing it separately, you can take advantage of pixel shift to increase detail in nonmoving areas, while masking out moving areas and painting over them with the single exposure. So you end up with pixel shift detail in the nonmoving areas, eliminating the motion artifacts while still retaining single-exposure detail in those areas.

Yes, I concur.  As far as I can see there is no way to do this automatically, however.  But we detail freaks can have at the stack and get some extraordinarily sharp landscapes.

Paul
www.PaulRoark.com

[eronald]

Michael,

 If you use Sony’s Remote app to trigger the camera the 4 pixelshiftedshots will be combined into a single ARQ file.

https://support.d-imaging.sony.co.jp/support/ilc/psms/en/index.html

Edmund

Thanks. I had previously read your article and learned from it. I have no trouble with understanding what pixel-shift is, since I used it extensively in the Pentax K3 and K1 cameras. Of course, it is frustrating that Adobe has not yet supported it, but I have the Imaging Edge software and the Sony free version of C1.

My problem, as primarily a focus-stacker, is how to use this technique in the A7R3 as part of my work process.

Did I understand what little there is from Sony about this technique that Imaging Edge will take one pixel-shift photo, figure out where all four of them are, and process that stack to deliver to us the combined photo? That would help. Then I have to turn each combined photo into a TIF file to send it to Zerene Stacker, my stacking software of choice. As you (I'm sure) already well understand, processes like these extend themselves until that is a lot of what we do. If I process a 100-layer stack, this a long process.

I do have a question, not about pixel-shifting, but how do I with Imagining Edge and C1.... batch process a group of A7R3 images, convert them to TIFF, and save them to a folder. I am on a PC. I have several Macs, but don't use them for this kind of work. PC has always been faster IMO, much less expensive, and I can tinker with them myself.

I am uncharacteristically grumpy about the menu system of the A7R3, although I had two copies of the A7R2, etc. Probably because in the last couple of years I have the Pentax K3 and K1, the Hasselblad X1D and the Fujifilm GFX, all of which I had to learn and work with. I am kind of tired of new cameras, but can't help myself but check them out. IMO, and this is just me, the A7R3 is like a toy (a good toy) compared to the Nikon D850, which I love. 

Shadowblade: the four pixel-shift images are raw and available like any other A7R3 images.

[Michael Erlewine]

Got that. I need to batch 50 sets of images all at once, to make it reasonable to do focus stacking, not one set at a time. And the resulting composite files need to be converted to TIF and saved. That idea.

[shadowblade]

I'm just wondering how you get natural subjects that aren't rocks to stay still for 100 low-ISO exposures without even moving a pixel, and how you get 100 images that line up perfectly at the pixel level despite shutter vibration.

Can the merged files be processed in other RAW converters (after all, they're still a kind of RAW file) or are you stuck using the Sony converter?