S1R pixel shifting

Have questions about the equipment used for macro- or micro- photography? Post those questions in this forum.

Moderators: rjlittlefield, ChrisR, Chris S., Pau

Lou Jost
Posts: 5944
Joined: Fri Sep 04, 2015 7:03 am
Location: Ecuador
Contact:

S1R pixel shifting

Post by Lou Jost »

Here is my first test of the Panasonic S1R pixel-shifting function. A quick review: this is the first full-frame camera to use 8-shot pixel shifting to quadruple the pixel dimensions of the sensor. This was previously implemented on Olympus and Panasonic Micro-Four-Thirds cameras. In that implementation, the pixel dimensions of the image were four times that of the sensor, but the effective resolution was about 2-2.5x that of the sensor. In this full-frame implementation, the nominal resolution of the image is 187Mp, but by analogy with the MFT case, the real resolution should be about 95-120Mp.

I don't have stacking equipment here in the US, so this is just a quick non-macro test. Here is the subject:

Image

Here is a crop of the middle tail feather. Left is unshifted, right is shifted. This is a tough subject, but a realistic one for nature photographers. The color moire of the unshifted image is extreme. There is absolutely no moire in the shifted image. There is also more resolution, and smoother background.

Image

People have sometimes expressed concern about false detail in pixel-shifted images. This image shows the opposite problem. The unshifted image has false detail due to moire, while the shifted image has only real detail, as far as I can see. Here is a detail (800% for the unshifted image, left, and 400% for the shifted image, right).

Image

Edited to add: This is taken with the Sigma 70mm Art lens in Canon mount on the Sigma MC21 EOS-L mount adapter, at f/5.6.

Edited again to note that I set sharpening to zero (not the default sharpening) in ACR. Additional sharpening improves the shifted images significantly, but makes the unshifted images worse.
Last edited by Lou Jost on Sun Jun 09, 2019 5:07 am, edited 3 times in total.

dmillard
Posts: 637
Joined: Tue Oct 24, 2006 7:37 pm
Location: Austin, Texas

Post by dmillard »

Thank you Lou - this is an impressive demonstration of the camera's potential!

Best regards,
David

mawyatt
Posts: 2497
Joined: Thu Aug 22, 2013 6:54 pm
Location: Clearwater, Florida

Post by mawyatt »

Lou,

Your test is really impressive and shows the benefit of the new Panasonic S1R pixel shift. How long does it take the camera to take the 8 individual images?

Best,
Research is like a treasure hunt, you don't know where to look or what you'll find!
~Mike

Lou Jost
Posts: 5944
Joined: Fri Sep 04, 2015 7:03 am
Location: Ecuador
Contact:

Post by Lou Jost »

Mike, there are four steps. The first is a user-selected trigger delay, which can be set to zero if desired. Then there is image capture, which seems instant, as fast as it takes to shoot eight shots. No write time to the card separates the shots; the results must go straight into a buffer. Then there is a processing step which takes about eleven seconds. Then there is the write stage, which depends on card speed and whether the user chooses to have a normal unshifted RAW file included in the output. On my 250Mb/s card this step (writing one shifted and one unshifted image) takes about ten seconds. A faster card might shorten this.

ray_parkhurst
Posts: 3413
Joined: Sat Nov 20, 2010 10:40 am
Location: Santa Clara, CA, USA
Contact:

Post by ray_parkhurst »

Very impressive, Lou!

I think the conclusion of the false detail discussion was that the compositing algorithm was sharpening the image and creating excessive micro-contrast that ended up making any small dark spot into a donut. This was on the A7Rii. Robert used some 3rd party software in his tests, and when I tried it the over-sharpening was gone and the image had "real" detail.

Can the S1R do tethered Live View on PC monitor, and direct download after compositing? How about triggering from a stacking controller like StackShot/WeMacro/mjkzz?

Lou Jost
Posts: 5944
Joined: Fri Sep 04, 2015 7:03 am
Location: Ecuador
Contact:

Post by Lou Jost »

Ray, there is a free dedicated tethering program for Panasonic cameras, Lumix Tether. I will set that up and test it when I get to Ecuador. I should add that Mike's custom controller can add flash synch to the "silent" shutter mode. This camera, like most others with the exception of Olympus, does not do flash with the electronic shutter.

One complaint about this camera, and it is a big one: the so-called partnership between Panasonic, Sigma, and Leica for the L mount lenses is not a full partnership; Sigma's EOS-to-L mount adapter, MC21, does not really make their Art lenses fully compatible with the camera, even though they are listed as "compatible". None of the advanced bracketing options are available for this camera.

I most wanted automatic focus bracketiing, like the Nikon Z. Though the camera was advertised as having focus bracketing, it appears never to have had the kind of focus bracketing that I had expected, and with the MC21 adapter, even the limited bracketing that does exist for native L lenses is not available. I should, however, be able to hack this with an Autohotkeys robot clicking on the tether app.

An alternative hack would be to use the WeMacro MicroMate to turn the lens focusing ring via a cine follow-focus gear device.

zzffnn
Posts: 1896
Joined: Thu May 22, 2014 1:25 pm
Location: Houston, Texas, USA
Contact:

Post by zzffnn »

Lou Jost wrote:.......Then there is image capture, which seems instant, as fast as it takes to shoot eight shots.......
Lou,

So that step #2 image capture step can be very fast, and maybe even finishes in 1/250 seconds (if each shot needs only 1/2000 seconds)?

There would be quite a few handhold-able niche applications, if that is the case.

I know pixel shifting from older Olympus is difficult to hand hold, but don't know about newer technologies from other camera brands.

mawyatt
Posts: 2497
Joined: Thu Aug 22, 2013 6:54 pm
Location: Clearwater, Florida

Post by mawyatt »

Lou Jost wrote:Ray, there is a free dedicated tethering program for Panasonic cameras, Lumix Tether. I will set that up and test it when I get to Ecuador. I should add that Mike's custom controller can add flash synch to the "silent" shutter mode. This camera, like most others with the exception of Olympus, does not do flash with the electronic shutter.

One complaint about this camera, and it is a big one: the so-called partnership between Panasonic, Sigma, and Leica for the L mount lenses is not a full partnership; Sigma's EOS-to-L mount adapter, MC21, does not really make their Art lenses fully compatible with the camera, even though they are listed as "compatible". None of the advanced bracketing options are available for this camera.

I most wanted automatic focus bracketiing, like the Nikon Z. Though the camera was advertised as having focus bracketing, it appears never to have had the kind of focus bracketing that I had expected, and with the MC21 adapter, even the limited bracketing that does exist for native L lenses is not available. I should, however, be able to hack this with an Autohotkeys robot clicking on the tether app.

An alternative hack would be to use the WeMacro MicroMate to turn the lens focusing ring via a cine follow-focus gear device.
Lou,

There are "extra channels" on the controller to allow all sorts of adaptations, 3 extra with the Pololu Tic-500s since they use USB (I know you could "expand" a USB channel for more channels, but not counting this option). So one could actually run the rail and focus stack with 2 Wemacro Micromates, and still have another two available.

When developing the Quad Axis controller I was thinking of using the 4th axis for direct focus control, so one could use the same setup and controller for moving the Z axis and/or camera focus control like the video folks have. Then someone mentioned the R axis, and started thinking where I abandoned this effort last year and realized I could drive the rotation axis with the same stepper motor!!

It may be possible to have a traditional 3 axis (XYZ) and Rotational axis and a focus control all from the same controller. Now the most I can do is 4 independent axis, but you've got me thinking about another development :shock:

As if I don't have enough going on already!! :roll:

Best,
Research is like a treasure hunt, you don't know where to look or what you'll find!
~Mike

mawyatt
Posts: 2497
Joined: Thu Aug 22, 2013 6:54 pm
Location: Clearwater, Florida

Post by mawyatt »

zzffnn wrote:
Lou Jost wrote:.......Then there is image capture, which seems instant, as fast as it takes to shoot eight shots.......
Lou,

So that step #2 image capture step can be very fast, and maybe even finishes in 1/250 seconds (if each shot needs only 1/2000 seconds)?

There would be quite a few handhold-able niche applications, if that is the case.

I know pixel shifting from older Olympus is difficult to hand hold, but don't know about newer technologies from other camera brands.
I'm not sure that this can be that fast since the writing to the buffer must occur after the Analog to Digital conversion of the pixels. Probably limited to the present high resolution camera rates of ~20ms or slower.

Best,
Research is like a treasure hunt, you don't know where to look or what you'll find!
~Mike

Lou Jost
Posts: 5944
Joined: Fri Sep 04, 2015 7:03 am
Location: Ecuador
Contact:

Post by Lou Jost »

I should put to rest any thoughts of handheld hi res shots. The camera can't even move 1 micron during the shooting sequence. Even on a tripod this is hard. Some day it may be possible to superimpose the sensor pixel shifting movements on the vibration reduction movements, but no camera does that now.

Lou Jost
Posts: 5944
Joined: Fri Sep 04, 2015 7:03 am
Location: Ecuador
Contact:

Post by Lou Jost »

Mike, that is great, the cine-type follow focus control will be really useful. Even without multiple channels, I could still just connect and disconnect the appropriate MicroMate to the controller, right?

palea
Posts: 40
Joined: Mon Dec 17, 2018 8:09 pm

Post by palea »

Lou Jost wrote:Some day it may be possible to superimpose the sensor pixel shifting movements on the vibration reduction movements, but no camera does that now.
Hi Lou, a closer look at the feature set of the Olympus E-M1X may be helpful. :wink:

It's also feasible to generate superresolution images by aligning frames from handheld bursts with offsets. Within conventional photography this is done informally (and practitioner implementations are often somewhat questionable) but it's fundamentally the same process as some of the more structured reconstructions relying on orbit to orbit variation in satellite imagery. Certain photographic tools, such as Google Night Sight and Kandao Raw+, perform operations close enough to single box pixel shifting (four exposures) to derive similar properties from sequences of handheld images. Nothing fundamentally prevents their extension to entry level superresolution implementations like the double box (eight exposures) high resolution modes currently available from Olympus and Panasonic.

(As an aside, focus brackets contain comparable information. To my knowledge, no stacker currently takes advantage of it for luminance or chroma noise reduction. While it would be computationally expensive, for higher ISO stacks I wouldn't mind the additional processing time. There are also a number of mechanisms, such as increased depth overlap between frames when using entocentric or pericentric optics, which would permit superresolution stacking.)

Lou Jost
Posts: 5944
Joined: Fri Sep 04, 2015 7:03 am
Location: Ecuador
Contact:

Post by Lou Jost »

Palea, that's impressive. It looks as though Olympus is doing what I thought could not yet be done! Panasonic does not make any such claimn for the S1R camera.

And yes, your point about getting super-resolution through multiple-shot averaging is a good one, and applies to any camera. Ray and I have both done some tests with this method. I think it is somewhat less efficient than structured shifting, but it is still helpful, as long as the subject is stationary (or if the moving areas are small or regular enough).

mawyatt
Posts: 2497
Joined: Thu Aug 22, 2013 6:54 pm
Location: Clearwater, Florida

Post by mawyatt »

Lou Jost wrote:Mike, that is great, the cine-type follow focus control will be really useful. Even without multiple channels, I could still just connect and disconnect the appropriate MicroMate to the controller, right?
Lou,

Yes. With the Trinamic based controller/driver PCB you have 3 "extra" available 2 pin power connectors. These are for powering Pololu Tic-500 Controller/Driver Motor Power, which interface with the Raspberry Pi via USB.

So you could use the Trinamic to control the normal Z axis with the WeMacro rail, or MicroMate, THK rail, or whatever and a Tic-500 to control another MicroMate which is mechanically linked to the lens focus control, or stepper motor linked to the focus control, all from the single Raspberry Pi!!

I'll post some images here if you want of the various controller/drivers under development if you want, but don't want to intrude on your thread without asking.

Best,
Research is like a treasure hunt, you don't know where to look or what you'll find!
~Mike

Lou Jost
Posts: 5944
Joined: Fri Sep 04, 2015 7:03 am
Location: Ecuador
Contact:

Post by Lou Jost »

Mike, I don't mind if you put photos on this page. Though I also think it is worth keeping a centralized dedicated page set for your controller.

Post Reply Previous topicNext topic