Pixel Shift with the Z8 - real world applications?

[PLShutterbug]

You mentioned hair as an example. They if, but even a four-image pixel shift takes almost a second so could be a problem with movement.

 

[Bruno Zarokian]

It seems to be taking about 100-120ms between frames (very rough test, someone may want to finesse), so 350-ish ms for 4 shots. Not too bad.

 

[Ernie Misner]

Hi Clint -

I just posted a first impressions review of pixel shift:

https://www.dpreview.com/forums/thread/4747192

My wife is an artist known for the incredible detail in her images. Especially for her larger work, having a lot of pixels helps when we print giclee prints. Even at 45MP I struggled to get the resolution she demands.

With pixel shift I can now get that. As the sample image of a portion of a map shows, the sharpening effect is dramatic.

So for reproduction work, pixel shift is very useful.

Thank you for the very useful comment. How did you process your images, and what are our options for processing them? Do you know if NX Studio has an option for that yet?

 

[PLShutterbug]

Hi Clint -

I just posted a first impressions review of pixel shift:

https://www.dpreview.com/forums/thread/4747192

My wife is an artist known for the incredible detail in her images. Especially for her larger work, having a lot of pixels helps when we print giclee prints. Even at 45MP I struggled to get the resolution she demands.

With pixel shift I can now get that. As the sample image of a portion of a map shows, the sharpening effect is dramatic.

So for reproduction work, pixel shift is very useful.

Thank you for the very useful comment. How did you process your images, and what are our options for processing them? Do you know if NX Studio has an option for that yet?

Hi Ernie-

NX Studio is the only app I know of for combining the different frames into a single image.

Once combined, a few apps can open the NEXF file that results.

 

[Ellis Vener]

Hi Clint -

I just posted a first impressions review of pixel shift:

https://www.dpreview.com/forums/thread/4747192

My wife is an artist known for the incredible detail in her images. Especially for her larger work, having a lot of pixels helps when we print giclee prints. Even at 45MP I struggled to get the resolution she demands.

With pixel shift I can now get that. As the sample image of a portion of a map shows, the sharpening effect is dramatic.

So for reproduction work, pixel shift is very useful.

Thank you for the very useful comment. How did you process your images, and what are our options for processing them? Do you know if NX Studio has an option for that yet?

Hi Ernie-

NX Studio is the only app I know of for combining the different frames into a single image.

Once combined, a few apps can open the NEXF file that results.

Or you can process the .NEXF file in NX Studio and then export a TIFF or JPEG to post-process in your favored post-processing program, including Lightroom.

 

[Ernie Misner]

Hi Clint -

I just posted a first impressions review of pixel shift:

https://www.dpreview.com/forums/thread/4747192

My wife is an artist known for the incredible detail in her images. Especially for her larger work, having a lot of pixels helps when we print giclee prints. Even at 45MP I struggled to get the resolution she demands.

With pixel shift I can now get that. As the sample image of a portion of a map shows, the sharpening effect is dramatic.

So for reproduction work, pixel shift is very useful.

Thank you for the very useful comment. How did you process your images, and what are our options for processing them? Do you know if NX Studio has an option for that yet?

Hi Ernie-

NX Studio is the only app I know of for combining the different frames into a single image.

Once combined, a few apps can open the NEXF file that results.

Thanks a bunch. Did you mean NEF or NEXF? So apparently just stacking the frames in PS does not work.

Edit - I see that NEXF is indeed a real file type. Is that still a raw file? I wonder what that stands for.

--
Ernie Misner
http://www.flickr.com/photos/erniemisner/
"The best photographer is the one who has the most fun!" - Jason Bradley, Outdoor Photographer Magazine

 

[PLShutterbug]

Hi Clint -

I just posted a first impressions review of pixel shift:

https://www.dpreview.com/forums/thread/4747192

My wife is an artist known for the incredible detail in her images. Especially for her larger work, having a lot of pixels helps when we print giclee prints. Even at 45MP I struggled to get the resolution she demands.

With pixel shift I can now get that. As the sample image of a portion of a map shows, the sharpening effect is dramatic.

So for reproduction work, pixel shift is very useful.

Thank you for the very useful comment. How did you process your images, and what are our options for processing them? Do you know if NX Studio has an option for that yet?

Hi Ernie-

NX Studio is the only app I know of for combining the different frames into a single image.

Once combined, a few apps can open the NEXF file that results.

Thanks a bunch. Did you mean NEF or NEXF? So apparently just stacking the frames in PS does not work.

Edit - I see that NEXF is indeed a real file type. Is that still a raw file? I wonder what that stands for.

Yes, I meant NEXF, and (my bad) actually the correct extension is NEFX. That's the RAW file type for a merged pixel-shift set that is saved as a single file by NX Studio.

Here is part of a directory listing showing a 32-shot pixel shift set with its NEFX file below, then a 4-shot set, then an 8-shot set. It looks pretty small in my preview - hopefully you can see it larger in Gallery view.

 

[Ryan Mack]

Hi Clint (and rest of the thread). I just posted a beta of my "Merge Raw" LR Classic Plugin (for macOS only right now) that lets you merge pixel shifted RAW files into a DNG in 2 clicks without leaving LR. I hope it offers both a much improved workflow, and in the case of images where something moved, better quality output that NX Studio. If you want to learn more I posted details on this other thread:

Lightroom plugin to combine pixel shift files?

The way my app works is it uses Adobe DNG converter to change all the NEFs to DNGs, then merges them into a single DNG either same pixel dimensions with lower noise and better color or into a larger DNG with higher res as well. That file gets added back into Lightroom for you. I'll try to get...

www.dpreview.com

 

[MoreorLess]

Whilst with a basic output ISO 64 files with the 45 MP sensor are essentially noiseless that can change if your doing heavy processing, especially if your say doing strong contrast adjustments on clouds with monochrome landscape shots.

 

[Ryan Mack]

Whilst with a basic output ISO 64 files with the 45 MP sensor are essentially noiseless that can change if your doing heavy processing, especially if your say doing strong contrast adjustments on clouds with monochrome landscape shots.

Yeah I notice because of the extra sharpening I like to do to improve prints. Noise in smooth shadow areas becomes visible on screen and on paper. That's actually what motivated me to work on my frame blending software originally. These days though Adobe's Enhanced Raw de-noising algorithm is so good at keeping details while removing noise from smooth areas it's not as important now.

 

[Ellis Vener]

Whilst with a basic output ISO 64 files with the 45 MP sensor are essentially noiseless that can change if your doing heavy processing, especially if your say doing strong contrast adjustments on clouds with monochrome landscape shots.

Yeah I notice because of the extra sharpening I like to do to improve prints. Noise in smooth shadow areas becomes visible on screen and on paper. That's actually what motivated me to work on my frame blending software originally. These days though Adobe's Enhanced Raw de-noising algorithm is so good at keeping details while removing noise from smooth areas it's not as important now.

And don't forget that in both Lightroom Classic, ACR, and the ACR filter in Photoshop, you can apply sharpening or noise reduction locally.

 

[panther fan]

So the biggest difference between NEFX and normal NEFs is in noise reduction at high ISO. However, I have to wonder when one would ever shoot at high ISO under the conditions required for Pixel Shift (no motion whatsoever in the scene, tripod) as I would normally shoot at the lowest ISO possible under these conditions.

As I responded to a similar question in another thread, if the light is changing it may take less time to shoot an 8-frame pixel shift sequence using a high ISO than to make a long exposure at at a low ISO. Think sunrise, sunset.

That's not how that works. A pixel shift will always be slower

 

[Ryan Mack]

Whilst with a basic output ISO 64 files with the 45 MP sensor are essentially noiseless that can change if your doing heavy processing, especially if your say doing strong contrast adjustments on clouds with monochrome landscape shots.

Yeah I notice because of the extra sharpening I like to do to improve prints. Noise in smooth shadow areas becomes visible on screen and on paper. That's actually what motivated me to work on my frame blending software originally. These days though Adobe's Enhanced Raw de-noising algorithm is so good at keeping details while removing noise from smooth areas it's not as important now.

And don't forget that in both Lightroom Classic, ACR, and the ACR filter in Photoshop, you can apply sharpening or noise reduction locally.

Sure. For printing I use Nik Output Sharpening and just blend the sharpened layer with the original layer. Used to I had to paint a mask but "select subject" generally does well enough and saves a lot of time. Still though I think Enhanced Raw does a better job of deciding on a per-pixel basis whether something should be sharpened or not.

 

[panther fan]

2. There is a very significant difference in noise reduction in all the NEFX files at ISO 1600 in comparison to the individual NEF files, even after using Adobe Denoise in ACR on the NEFs.

3. I could see no difference in noise at ISO 64 between the NEFX and the NEFs. The Z8 NEFs are really clean at ISO 64.

As Horshack wrote, the difference in noise is actually the same (about a half stop for 4x, about a full stop for 8x). What you're really writing here is that the noise was buried in deep shadow at ISO 64, but visible in mid-tones at ISO 1600.

Ok thanks for that. I understand. I was really thinking about situations where one would be in control of the lighting and exposures would be in the range where digging down into the deep shadows would not be necessary and normal NEFs would generally do the trick.

Even if they don,t just regular frame blending or a simple bracket will do the same trick. No need for pixel shift.

Pixel shift is really only usefull for resolution/ artefacts

 

[Ryan Mack]

So the biggest difference between NEFX and normal NEFs is in noise reduction at high ISO. However, I have to wonder when one would ever shoot at high ISO under the conditions required for Pixel Shift (no motion whatsoever in the scene, tripod) as I would normally shoot at the lowest ISO possible under these conditions.

As I responded to a similar question in another thread, if the light is changing it may take less time to shoot an 8-frame pixel shift sequence using a high ISO than to make a long exposure at at a low ISO. Think sunrise, sunset.

That's not how that works. A pixel shift will always be slower

Yeah pixel shift would capture the same amount of light but with tiny pauses between images so total time would be increased. For a stationary subject noise would be lower with pixel shift because it reduces some artifacts caused by demosaicing. For a moving subject noise would be higher because you would need to throw out some pixels to avoid artifacts (or keep the artifacts for extra noise).

Once nice thing about frame averaging (pixel shift or not) is that you can determine the exposure length later. So if you have a waterfall or stream, shoot a set of 30 .1 second exposures and you can decide later how many to blend to get the amount of blur you like. You might get some artifacts caused by the gaps between exposures but for slower shutter speeds I don't think it would be very visible.

 

[panther fan]

So the biggest difference between NEFX and normal NEFs is in noise reduction at high ISO. However, I have to wonder when one would ever shoot at high ISO under the conditions required for Pixel Shift (no motion whatsoever in the scene, tripod) as I would normally shoot at the lowest ISO possible under these conditions.

As I responded to a similar question in another thread, if the light is changing it may take less time to shoot an 8-frame pixel shift sequence using a high ISO than to make a long exposure at at a low ISO. Think sunrise, sunset.

That's not how that works. A pixel shift will always be slower

Yeah pixel shift would capture the same amount of light but with tiny pauses between images so total time would be increased. For a stationary subject noise would be lower with pixel shift because it reduces some artifacts caused by demosaicing. For a moving subject noise would be higher because you would need to throw out some pixels to avoid artifacts (or keep the artifacts for extra noise).

Once nice thing about frame averaging (pixel shift or not) is that you can determine the exposure length later. So if you have a waterfall or stream, shoot a set of 30 .1 second exposures and you can decide later how many to blend to get the amount of blur you like. You might get some artifacts caused by the gaps between exposures but for slower shutter speeds I don't think it would be very visible.

That would cause severe debayering artefacts (also visible on all other brands pixel shift)

You then would have to do two things:

1. Develop the image pixel shifted for the stationary parts

2. Develop each frame seperately, mean merge all the files, then paint in the areas where you have motion. (you will only get non pixel shifted resolution, but that doesn't matter for a long exposure effect of movement)

 

[Ryan Mack]

2. There is a very significant difference in noise reduction in all the NEFX files at ISO 1600 in comparison to the individual NEF files, even after using Adobe Denoise in ACR on the NEFs.

3. I could see no difference in noise at ISO 64 between the NEFX and the NEFs. The Z8 NEFs are really clean at ISO 64.

As Horshack wrote, the difference in noise is actually the same (about a half stop for 4x, about a full stop for 8x). What you're really writing here is that the noise was buried in deep shadow at ISO 64, but visible in mid-tones at ISO 1600.

Ok thanks for that. I understand. I was really thinking about situations where one would be in control of the lighting and exposures would be in the range where digging down into the deep shadows would not be necessary and normal NEFs would generally do the trick.

Even if they don,t just regular frame blending or a simple bracket will do the same trick. No need for pixel shift.

Pixel shift is really only usefull for resolution/ artefacts

I don't really think that's true. If you frame blend 8 photos at the same pixel location you end up with uneven noise across the colors. The DNG headers include the per-channel noise characteristic which I use to compute the motion artifact threshold in my frame blending software. With pixel shift you end up with a uniform distribution of noise across the colors. I think that should result in overall lower noise than in the non-shifted image. Again, assuming truly stationary camera and subject.

 

[Ryan Mack]

So the biggest difference between NEFX and normal NEFs is in noise reduction at high ISO. However, I have to wonder when one would ever shoot at high ISO under the conditions required for Pixel Shift (no motion whatsoever in the scene, tripod) as I would normally shoot at the lowest ISO possible under these conditions.

As I responded to a similar question in another thread, if the light is changing it may take less time to shoot an 8-frame pixel shift sequence using a high ISO than to make a long exposure at at a low ISO. Think sunrise, sunset.

That's not how that works. A pixel shift will always be slower

Yeah pixel shift would capture the same amount of light but with tiny pauses between images so total time would be increased. For a stationary subject noise would be lower with pixel shift because it reduces some artifacts caused by demosaicing. For a moving subject noise would be higher because you would need to throw out some pixels to avoid artifacts (or keep the artifacts for extra noise).

Once nice thing about frame averaging (pixel shift or not) is that you can determine the exposure length later. So if you have a waterfall or stream, shoot a set of 30 .1 second exposures and you can decide later how many to blend to get the amount of blur you like. You might get some artifacts caused by the gaps between exposures but for slower shutter speeds I don't think it would be very visible.

That would cause severe debayering artefacts (also visible on all other brands pixel shift)

You then would have to do two things:

1. Develop the image pixel shifted for the stationary parts

2. Develop each frame seperately, mean merge all the files, then paint in the areas where you have motion. (you will only get non pixel shifted resolution, but that doesn't matter for a long exposure effect of movement)

Exactly! I just spent the last week making a Lightroom Plugin that does exactly that! But much faster than if you did it manually in Photoshop.

https://www.dpreview.com/forums/post/67539396

Edit: To clarify, in the non-pixel shifted case you can do the arbitrary blending for extending blur across multiple images. In the pixel shifted case right now you can only do a single set of pixel shifted images. I like the idea of allowing multiple sets of pixel shifted images (with motion artifact removal) and then trying to merge the resulting files into a single longer exposure. I'll add that to the roadmap.

 

[panther fan]

2. There is a very significant difference in noise reduction in all the NEFX files at ISO 1600 in comparison to the individual NEF files, even after using Adobe Denoise in ACR on the NEFs.

3. I could see no difference in noise at ISO 64 between the NEFX and the NEFs. The Z8 NEFs are really clean at ISO 64.

As Horshack wrote, the difference in noise is actually the same (about a half stop for 4x, about a full stop for 8x). What you're really writing here is that the noise was buried in deep shadow at ISO 64, but visible in mid-tones at ISO 1600.

Ok thanks for that. I understand. I was really thinking about situations where one would be in control of the lighting and exposures would be in the range where digging down into the deep shadows would not be necessary and normal NEFs would generally do the trick.

Even if they don,t just regular frame blending or a simple bracket will do the same trick. No need for pixel shift.

Pixel shift is really only usefull for resolution/ artefacts

I don't really think that's true. If you frame blend 8 photos at the same pixel location you end up with uneven noise across the colors. The DNG headers include the per-channel noise characteristic which I use to compute the motion artifact threshold in my frame blending software. With pixel shift you end up with a uniform distribution of noise across the colors. I think that should result in overall lower noise than in the non-shifted image. Again, assuming truly stationary camera and subject.

?

Post debayer, it would still be fine, but that takes to long to explain.

So much simpler (both in software and in explanation), just blend the RAW images pre-debayer. This way you get pretty much the capture with a theoretical sensor with higher FWC and just much longer exposure. You cannot get better than that (in terms of noise) with pixel shift

 

[Ryan Mack]

2. There is a very significant difference in noise reduction in all the NEFX files at ISO 1600 in comparison to the individual NEF files, even after using Adobe Denoise in ACR on the NEFs.

3. I could see no difference in noise at ISO 64 between the NEFX and the NEFs. The Z8 NEFs are really clean at ISO 64.

As Horshack wrote, the difference in noise is actually the same (about a half stop for 4x, about a full stop for 8x). What you're really writing here is that the noise was buried in deep shadow at ISO 64, but visible in mid-tones at ISO 1600.

Ok thanks for that. I understand. I was really thinking about situations where one would be in control of the lighting and exposures would be in the range where digging down into the deep shadows would not be necessary and normal NEFs would generally do the trick.

Even if they don,t just regular frame blending or a simple bracket will do the same trick. No need for pixel shift.

Pixel shift is really only usefull for resolution/ artefacts

I don't really think that's true. If you frame blend 8 photos at the same pixel location you end up with uneven noise across the colors. The DNG headers include the per-channel noise characteristic which I use to compute the motion artifact threshold in my frame blending software. With pixel shift you end up with a uniform distribution of noise across the colors. I think that should result in overall lower noise than in the non-shifted image. Again, assuming truly stationary camera and subject.

?

Post debayer, it would still be fine, but that takes to long to explain.

So much simpler (both in software and in explanation), just blend the RAW images pre-debayer. This way you get pretty much the capture with a theoretical sensor with higher FWC and just much longer exposure. You cannot get better than that (in terms of noise) with pixel shift

Agreed. My LR plugin outputs a pre-debayer image if you are not using pixel-shift or single-pixel alignment do not enable the de-ghosting mode (which necessitates demosaicing for comparison and to fill in images with movement beyond calculated noise threshold).

That was actually the original motivation for writing it. I was tired of using frame averaging in photoshop and creating massive demosaiced image files. Merging pre-debayer images is super fast and the resulting DNG losslessly compresses better than the original because of lower noise.