K3II: Pixel Shift First Test

**lm4187** · 05-27-2015, 07:25 AM

Originally posted by Barry Pearson

My main complaint is that they contravened the DNG specification. That is doomed to cause problems.

How so?

Cheers,
Lou

DeadJohn · 05-27-2015, 08:35 AM

I hope Ricoh releases definitive recommendations about how to best use pixel shift. Every outdoor sample I've looked at shows artifacts due to movement. How can artifacts be removed without manually searching for and fixing each occurrence in Photoshop? Applying a blur filter to the entire image would undo pixel shift's gains. The pixellation is showing in JPGs so it's not just a raw processing issue.

Ricoh might also need to provide guidance, or update firmware, or work with Adobe so pixel shifted DNG files can be loaded directly into Lightroom.

A question for the early adopters: Can you run a test to see if there's a processing delay between each subframe or just at the end? A delay between subframes increases the risk of movement artifacts. Maybe set an extreme case of 1 second exposure with pixel shift on. Does it take exactly 4 seconds to finish the last exposure?

Gimbal · 05-27-2015, 09:00 AM

A "fun" picture would be a PS shot of a watch with a moving (preferably a sweeping) second hand, done with a short exposure. That should give us a hint of time between shots.

Barry Pearson · 05-27-2015, 09:24 AM

Originally posted by lm4187

How so?

Do you mean "apart from the fact that software that can handle proper DNG can't handle these DNGs"?

In these forums and elsewhere a lot of time has been wasted. There is the risk that there are false expectations about the capability of this camera.

And, as a computer systems engineer before I retired, helping to design complex multi-vendor computer systems, I know the importance of establishing clear interfaces and sticking to them.

tomtor · 05-27-2015, 10:14 AM

Originally posted by Barry Pearson

Do you mean "apart from the fact that software that can handle proper DNG can't handle these DNGs"?

In these forums and elsewhere a lot of time has been wasted. There is the risk that there are false expectations about the capability of this camera.

And, as a computer systems engineer before I retired, helping to design complex multi-vendor computer systems, I know the importance of establishing clear interfaces and sticking to them.

In the strict sense it IS a conforming DNG (nothing forbids storing 4 CFA images in a single DNG) with the bonus I mentioned.

Other approaches are possible (eg writing a DNG and a second Tiff/linear-DNG with the PS data) but that is also not ideal.

Your preference for just a single linear DNG would probably cause more issues for many existing raw development programs.

The approach Ricoh has choosen is a reasonable one IMO. They are not idiots.

eraser · 05-27-2015, 10:16 AM

Originally posted by DeadJohn

I hope Ricoh releases definitive recommendations about how to best use pixel shift. Every outdoor sample I've looked at shows artifacts due to movement. How can artifacts be removed without manually searching for and fixing each occurrence in Photoshop? Applying a blur filter to the entire image would undo pixel shift's gains. The pixellation is showing in JPGs so it's not just a raw processing issue.

Ricoh might also need to provide guidance, or update firmware, or work with Adobe so pixel shifted DNG files can be loaded directly into Lightroom.

A question for the early adopters: Can you run a test to see if there's a processing delay between each subframe or just at the end? A delay between subframes increases the risk of movement artifacts. Maybe set an extreme case of 1 second exposure with pixel shift on. Does it take exactly 4 seconds to finish the last exposure?

Pretty sure the processing all happens after the exposures as there is a "processing" progress bar dialog that is displayed after the shutter closes.

Matthew Saville · 05-27-2015, 10:26 AM

It seems they are handling artifacts / ghosting the way I had hoped we were past at this point. When faced with movement, the pixel-shift system should simply revert to one of the lone exposures for that whole area. I'd rather see a slight decrease in acuity in some unimportant detail, than all kinds of horrible ghosts / artifacts.

I was actually rather impressed with how the K-3 originally handled artifacts in their in-camera HDR system, but this seems to fall short of what I expected from a natural progression.

Having said that, it's a new feature, a 1.0 version at that, and so I don't feel disappointed. It'll surely be improved, either in the next camera, or maybe even via firmware. All in all, I'm much more excited about the K-3 II for other reasons. :-)

tomtor · 05-27-2015, 10:39 AM

Originally posted by Matthew Saville

It seems they are handling artifacts / ghosting the way I had hoped we were past at this point. When faced with movement, the pixel-shift system should simply revert to one of the lone exposures for that whole area. I'd rather see a slight decrease in acuity in some unimportant detail, than all kinds of horrible ghosts / artifacts.

That would be nice (automatic artifact detection) but it is far from simple.

The PS is nice but you'll have to adapt your workflow to it or buy a 645

melander · 05-27-2015, 10:41 AM

was it not communicated that it would only work in pentax own raw developer. Pretty sure i read it beforehand.

tomtor · 05-27-2015, 10:54 AM

Originally posted by eraser

Pretty sure the processing all happens after the exposures as there is a "processing" progress bar dialog that is displayed after the shutter closes.

As you are one of the first with a k3ii, have you already looked at the Pentax supplied raw software?

Does it have an export option for DNG or TIFF (16 bit)?

JimmyDranox · 05-27-2015, 11:00 AM

IMO, Ricoh has start a good job with that pixel shift, but more work is needed.

I have seen a better DR, and more resolution in those images, but also some artifacts at pixel level, and also some strange color shift. Hopefully, everything can be fixed with a firmware.

I will post some pictures, at bigger than 100%, to illustrate what I have said.

In the first picture, at 500%. At left is the pixel shift On.

Second picture, 200%, at left pixel shift On.

Third image, a little more purple fringe than w/o pixel shift. And noise on the windows is more obvious with pixel shift.

JimmyDranox · 05-27-2015, 11:40 AM

Our fellow member ''eraser'' has posted some very good picture with K-3II and with pixel shift, and I took the permission to make a comparison with his images.

I think that's a good example of the gain in DR and resolution which can be achieved with this new technology.

I must say that the difference is much bigger on my TNT monitor, that on the laptop LED screen. So, I will ad a 150% screen capture. PS On is on the right.

eraser · 05-27-2015, 12:12 PM

Originally posted by tomtor

As you are one of the first with a k3ii, have you already looked at the Pentax supplied raw software?

Does it have an export option for DNG or TIFF (16 bit)?

I can check that tonight. The only manual or software I've looked at was the included flu card (that was a nice surprise). Only used my normal Lightroom workflow.

Originally posted by JimmyDranox

Our fellow member ''eraser'' has posted some very good picture with K-3II and with pixel shift, and I took the permission to make a comparison with his images.

I think that's a good example of the gain in DR and resolution which can be achieved with this new technology.

I must say that the difference is much bigger on my TNT monitor, that on the laptop LED screen. So, I will ad a 150% screen capture. PS On is on the right.

My impression is that ps gives a huge increase in detail on jpgs out of the camera and when developing dngs I see an even larger amount of detail. Even with good lenses such as DA* 16-50/50-135 and sigma 70-200 f2.8 I'm seeing a big difference in detail before doing any developing/sharpening. I would shoot with it on most of the time if I could.
Going to play with some handheld ps shots tonight and see how it behaves with high shutter speeds.

Barry Pearson · 05-27-2015, 12:13 PM - **1 Like**

Originally posted by tomtor

In the strict sense it IS a conforming DNG (nothing forbids storing 4 CFA images in a single DNG) with the bonus I mentioned.

What part of the DNG specification permits that?
Section 2 says "The highest-resolution and quality IFD should use NewSubFileType equal to 0. Reduced resolution (or quality) thumbnails or previews, if any, should use NewSubFileType equal to 1 (for a primary preview) or 10001.H (for an alternate preview)".
Note "The ...".

Originally posted by tomtor

Your preference for just a single linear DNG would probably cause more issues for many existing raw development programs.

Why? We know that all Adobe products support it! My investigation years ago is that several other raw converters supported it too:
Products supporting Linear DNG at various times between 2006 and 2010.

Originally posted by tomtor

The approach Ricoh has choosen is a reasonable one IMO. They are not idiots.

They've used DNG in a way that Adobe's own products can't handle properly! Lightroom! ACR in Photoshop!
Idiots!

rawr · 05-27-2015, 02:23 PM

Originally posted by JimmyDranox

a little more purple fringe than w/o pixel shift

It seems unlikely that, by itself, pixel shifting would correct purple fringing.

From reading the manual, pixel shifting can work with the in-camera corrections for distortion correction, CA and vignetting. Maybe those corrections were turned on here.

But then again, if those corrections were turned on, the first non-pixel shifted JPEG shouldn't have had any PF in it either. Hmmm. Interesting.

05-27-2015, 10:14 AM	#80
tomtor Veteran Member Join Date: Apr 2008 Location: Netherlands Posts: 382	Originally posted by Barry Pearson Do you mean "apart from the fact that software that can handle proper DNG can't handle these DNGs"? In these forums and elsewhere a lot of time has been wasted. There is the risk that there are false expectations about the capability of this camera. And, as a computer systems engineer before I retired, helping to design complex multi-vendor computer systems, I know the importance of establishing clear interfaces and sticking to them. In the strict sense it IS a conforming DNG (nothing forbids storing 4 CFA images in a single DNG) with the bonus I mentioned. Other approaches are possible (eg writing a DNG and a second Tiff/linear-DNG with the PS data) but that is also not ideal. Your preference for just a single linear DNG would probably cause more issues for many existing raw development programs. The approach Ricoh has choosen is a reasonable one IMO. They are not idiots. Last edited by tomtor; 05-27-2015 at 10:32 AM. Reason: typo

05-27-2015, 11:00 AM	#86
JimmyDranox Veteran Member Join Date: Oct 2012 Location: Ploiesti, Romania Photos: Gallery \| Albums Posts: 1,632	IMO, Ricoh has start a good job with that pixel shift, but more work is needed. I have seen a better DR, and more resolution in those images, but also some artifacts at pixel level, and also some strange color shift. Hopefully, everything can be fixed with a firmware. I will post some pictures, at bigger than 100%, to illustrate what I have said. In the first picture, at 500%. At left is the pixel shift On. Second picture, 200%, at left pixel shift On. Third image, a little more purple fringe than w/o pixel shift. And noise on the windows is more obvious with pixel shift. Last edited by JimmyDranox; 04-17-2017 at 10:26 PM.

05-27-2015, 11:40 AM	#87
JimmyDranox Veteran Member Join Date: Oct 2012 Location: Ploiesti, Romania Photos: Gallery \| Albums Posts: 1,632	Our fellow member ''eraser'' has posted some very good picture with K-3II and with pixel shift, and I took the permission to make a comparison with his images. I think that's a good example of the gain in DR and resolution which can be achieved with this new technology. I must say that the difference is much bigger on my TNT monitor, that on the laptop LED screen. So, I will ad a 150% screen capture. PS On is on the right. Last edited by JimmyDranox; 04-17-2017 at 10:26 PM.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
SR and Pixel Shift Technology	C_Jones	Pentax K-3 & K-3 II	9	05-10-2015 06:24 AM
Pixel Shifting Samples from Pentax on the K3II	mikeodial	Pentax DSLR Discussion	99	04-26-2015 05:00 PM
Pixel shift practical applications	Spodeworld	Pentax K-3 & K-3 II	20	04-25-2015 10:01 AM
K-3 II pixel shift resized. Wow factor.	JimmyDranox	Pentax K-3 & K-3 II	21	04-25-2015 09:37 AM
Pixel shift and bracketing	Spodeworld	Pentax K-3 & K-3 II	9	04-23-2015 08:56 PM

05-27-2015, 07:25 AM	#76
lm4187 Site Supporter Join Date: Jul 2009 Location: Germantown MD Posts: 90	Originally posted by Barry Pearson My main complaint is that they contravened the DNG specification. That is doomed to cause problems. How so? Cheers, Lou

05-27-2015, 08:35 AM	#77
DeadJohn Pentaxian Join Date: Jan 2011 Location: New York Posts: 4,834	I hope Ricoh releases definitive recommendations about how to best use pixel shift. Every outdoor sample I've looked at shows artifacts due to movement. How can artifacts be removed without manually searching for and fixing each occurrence in Photoshop? Applying a blur filter to the entire image would undo pixel shift's gains. The pixellation is showing in JPGs so it's not just a raw processing issue. Ricoh might also need to provide guidance, or update firmware, or work with Adobe so pixel shifted DNG files can be loaded directly into Lightroom. A question for the early adopters: Can you run a test to see if there's a processing delay between each subframe or just at the end? A delay between subframes increases the risk of movement artifacts. Maybe set an extreme case of 1 second exposure with pixel shift on. Does it take exactly 4 seconds to finish the last exposure?

05-27-2015, 09:00 AM	#78
Gimbal Pentaxian Join Date: Apr 2007 Location: Sweden Posts: 2,106	A "fun" picture would be a PS shot of a watch with a moving (preferably a sweeping) second hand, done with a short exposure. That should give us a hint of time between shots.

05-27-2015, 09:24 AM	#79
Barry Pearson Veteran Member Join Date: Jul 2013 Location: Stockport Posts: 964	Originally posted by lm4187 How so? Do you mean "apart from the fact that software that can handle proper DNG can't handle these DNGs"? In these forums and elsewhere a lot of time has been wasted. There is the risk that there are false expectations about the capability of this camera. And, as a computer systems engineer before I retired, helping to design complex multi-vendor computer systems, I know the importance of establishing clear interfaces and sticking to them.

05-27-2015, 10:16 AM	#81
eraser Junior Member Join Date: Jun 2007 Location: Victoria, BC Posts: 32	Originally posted by DeadJohn I hope Ricoh releases definitive recommendations about how to best use pixel shift. Every outdoor sample I've looked at shows artifacts due to movement. How can artifacts be removed without manually searching for and fixing each occurrence in Photoshop? Applying a blur filter to the entire image would undo pixel shift's gains. The pixellation is showing in JPGs so it's not just a raw processing issue. Ricoh might also need to provide guidance, or update firmware, or work with Adobe so pixel shifted DNG files can be loaded directly into Lightroom. A question for the early adopters: Can you run a test to see if there's a processing delay between each subframe or just at the end? A delay between subframes increases the risk of movement artifacts. Maybe set an extreme case of 1 second exposure with pixel shift on. Does it take exactly 4 seconds to finish the last exposure? Pretty sure the processing all happens after the exposures as there is a "processing" progress bar dialog that is displayed after the shutter closes.

05-27-2015, 10:26 AM	#82
Matthew Saville Senior Member Join Date: Jan 2014 Location: Southern California Posts: 250	It seems they are handling artifacts / ghosting the way I had hoped we were past at this point. When faced with movement, the pixel-shift system should simply revert to one of the lone exposures for that whole area. I'd rather see a slight decrease in acuity in some unimportant detail, than all kinds of horrible ghosts / artifacts. I was actually rather impressed with how the K-3 originally handled artifacts in their in-camera HDR system, but this seems to fall short of what I expected from a natural progression. Having said that, it's a new feature, a 1.0 version at that, and so I don't feel disappointed. It'll surely be improved, either in the next camera, or maybe even via firmware. All in all, I'm much more excited about the K-3 II for other reasons. :-)

05-27-2015, 10:39 AM	#83
tomtor Veteran Member Join Date: Apr 2008 Location: Netherlands Posts: 382	Originally posted by Matthew Saville It seems they are handling artifacts / ghosting the way I had hoped we were past at this point. When faced with movement, the pixel-shift system should simply revert to one of the lone exposures for that whole area. I'd rather see a slight decrease in acuity in some unimportant detail, than all kinds of horrible ghosts / artifacts. That would be nice (automatic artifact detection) but it is far from simple. The PS is nice but you'll have to adapt your workflow to it or buy a 645

05-27-2015, 10:41 AM	#84
melander Veteran Member Join Date: Jul 2008 Location: Sweden, Umea Posts: 876	was it not communicated that it would only work in pentax own raw developer. Pretty sure i read it beforehand.

05-27-2015, 10:54 AM	#85
tomtor Veteran Member Join Date: Apr 2008 Location: Netherlands Posts: 382	Originally posted by eraser Pretty sure the processing all happens after the exposures as there is a "processing" progress bar dialog that is displayed after the shutter closes. As you are one of the first with a k3ii, have you already looked at the Pentax supplied raw software? Does it have an export option for DNG or TIFF (16 bit)?

05-27-2015, 12:12 PM	#88
eraser Junior Member Join Date: Jun 2007 Location: Victoria, BC Posts: 32	Originally posted by tomtor As you are one of the first with a k3ii, have you already looked at the Pentax supplied raw software? Does it have an export option for DNG or TIFF (16 bit)? I can check that tonight. The only manual or software I've looked at was the included flu card (that was a nice surprise). Only used my normal Lightroom workflow. Originally posted by JimmyDranox Our fellow member ''eraser'' has posted some very good picture with K-3II and with pixel shift, and I took the permission to make a comparison with his images. I think that's a good example of the gain in DR and resolution which can be achieved with this new technology. I must say that the difference is much bigger on my TNT monitor, that on the laptop LED screen. So, I will ad a 150% screen capture. PS On is on the right. My impression is that ps gives a huge increase in detail on jpgs out of the camera and when developing dngs I see an even larger amount of detail. Even with good lenses such as DA* 16-50/50-135 and sigma 70-200 f2.8 I'm seeing a big difference in detail before doing any developing/sharpening. I would shoot with it on most of the time if I could. Going to play with some handheld ps shots tonight and see how it behaves with high shutter speeds.

05-27-2015, 12:13 PM - 1 Like	#89
Barry Pearson Veteran Member Join Date: Jul 2013 Location: Stockport Posts: 964	Originally posted by tomtor In the strict sense it IS a conforming DNG (nothing forbids storing 4 CFA images in a single DNG) with the bonus I mentioned. What part of the DNG specification permits that? Section 2 says "The highest-resolution and quality IFD should use NewSubFileType equal to 0. Reduced resolution (or quality) thumbnails or previews, if any, should use NewSubFileType equal to 1 (for a primary preview) or 10001.H (for an alternate preview)". Note "The ...". Originally posted by tomtor Your preference for just a single linear DNG would probably cause more issues for many existing raw development programs. Why? We know that all Adobe products support it! My investigation years ago is that several other raw converters supported it too: Products supporting Linear DNG at various times between 2006 and 2010. Originally posted by tomtor The approach Ricoh has choosen is a reasonable one IMO. They are not idiots. They've used DNG in a way that Adobe's own products can't handle properly! Lightroom! ACR in Photoshop! Idiots!

05-27-2015, 02:23 PM	#90
rawr Veteran Member Join Date: Jun 2009 Posts: 11,913	Originally posted by JimmyDranox a little more purple fringe than w/o pixel shift It seems unlikely that, by itself, pixel shifting would correct purple fringing. From reading the manual, pixel shifting can work with the in-camera corrections for distortion correction, CA and vignetting. Maybe those corrections were turned on here. But then again, if those corrections were turned on, the first non-pixel shifted JPEG shouldn't have had any PF in it either. Hmmm. Interesting.