The GetDPI Photography Forum

Great to see you here. Join our insightful photographic forum today and start tapping into a huge wealth of photographic knowledge. Completing our simple registration process will allow you to gain access to exclusive content, add your own topics and posts, share your work and connect with other members through your own private inbox! And don’t forget to say hi!

Fujifilm GFX100 Pixelshift Firmware is out

Paul2660

Well-known member
What is so crazy about this is that's quite apparent that in most normal conditions, you really can't use this tool. It appears you need a solid metal floor, or concrete. The fact in the video, that you are told to turn off the sounds the camera makes, as they create resonate inside the camera, and you can't use a remote as your own foot falls on the floor moving away from the camera effect the shot. I was totally incorrect in my assumption that the IBIS unit would not only allow for the shift, but would also help to compensate for the vibration issues. It's pretty clear to me that in most shooting environments, this feature is really moot.

Paul C
 

f8orbust

Member
I think the feature highlights something that's been there all along - when you have pixels (sensels?) as small as a few 1000ths of a mm, it's impossible (outside of a lab) to stop the camera moving at least this much during an exposure (even with a relatively fast shutter speed and solid tripod) and thus reducing the effective MP count of your camera. Pixelshifting without some sort of compensatory mechanism, e.g. IS in the lens, just puts a spotlight on this problem - one which, hitherto, has gone under the radar somewhat. Maybe this is why many folks like the look of 'fat pixel' digital backs, since they really are capturing (or getting closer to capturing) what it says on the can in terms of MP :unsure:
 
Last edited:
Im so happy i found this thread because I thought i was craZy and my gfx100 was faulty when i shot my pixel-shift images and the results were horrible :)
 
Btw i got somw good results i will post soon shooting on a tripod, profoto strobe, gf45mm. Amazing results despite i live 5th floor hardwood floor in Manhattan where vibrations are constant all day.
I dont believe the theory you need a bunker.
Also my older Sony a7r3 was great at pixel shift wit no need of all these attentions /cautions
 
consider the thread spool here was just a small element of the entire image shot with GF45mm
unfortunately, I lost the original image for comparison, but you can see how much more detail and less noise is the pixel-shift image in the center

Screen Shot 2020-12-03 at 12.43.58 PM.jpg
 

f8orbust

Member
You definitely don't need a bunker - you could shoot it handheld on a fairground ride if you want - but the less movement, the better the results will be.

Your photos clearly show an improvement over a single image (that said, the uprezzed image isn't a million miles away - smooth out the noise, get rid of the halos and sharpen it appropriately and it would be very close), but in terms of resolution it's not as an impressive a (relative) 'jump' as I've seen between Olympus hi-res shots and the uprezzed (Olympus) originals (though of course both are much lower resolution than what's possible with the 100MP Fuji), which again leads me to think that movement is an issue.

While I'm sure Fuji has some clever averaging going on in software to help deal with the issue, I still stand by the notion that it's pretty much impossible to stop the camera moving by a few 1000ths of a mm during one exposure, let alone 16, and this movement has to impact the final image. Years ago I bolted down a Leica laser to a huge Foba studio stand and pointed it at a thin frosted piece of plastic a few feet away. On the other side of the plastic I set up a camera on a studio stand and filmed the dot. When I played the video back I could see the dot moving (not much, but enough). Or was it the camera moving ? Or both ? You see the problem. Since then I've always taken the MP claim of a camera manufacturer with a pinch of salt (as in, if it says 50MP you're obviously going to get 50MP, but what those pixels actually represent is a bit up in the air - more so if it's a windy day).

All said and done, it's impossible to precisely quantify outside of a lab, so it's something of a moot point in terms of practical day-to-day use. Ultimately, if the photographer is happy with the image that's all that matters.
 
Last edited:

rga

Member
I'm thinking that the pixel shift function CAN be useful for landscapes that do not have any movement (e.g., Cholla cactus in Joshua tree, Death Valley Canyons). Waterfall landscapes without foliage (Iceland?) may also work, though aliasing on the moving water would be interesting to test, esp. at slow shutter speeds. Same with a stationery object in water; slower shutter speeds.

Finally, I have read where Helicon Focus provides a better platform for stacking the 16 images than the Fuji SW, but I don't know if the raw files produced can be processed. Perhaps bringing the files into C1 or LR, developing the raw files to taste (actually developing one and applying those adjustments to all 16 files), and then stacking in Helicon Focus may be the way to go.

I'm not sure about cloudy skies at slow shutter speeds. Just need to test. My gut says no, but can't think of a logical reason...
Any thoughts?
 
I went out for some city shots, carbon tripod, GF45mm and slow speed (1/4 or so) because it was late afternoon and man the results were horrible. horrible. trashed everything.
It wasn't windy. i probably did something wrong idk...

With my Sony a7r3 (only 4-shot mode) i don't remember to be so difficult to get good results jeezus...it was so easy...

ALSO
I saw the video of the guy affirming you have use 2s or longer intervals... well I will post a test soon, but to me the best results were with 0 is interval...

I feel the FUJI software needs improvement... I cant believe the results is so random... ugh
 
I'm thinking that the pixel shift function CAN be useful for landscapes that do not have any movement (e.g., Cholla cactus in Joshua tree, Death Valley Canyons). Waterfall landscapes without foliage (Iceland?) may also work, though aliasing on the moving water would be interesting to test, esp. at slow shutter speeds. Same with a stationery object in water; slower shutter speeds.

Finally, I have read where Helicon Focus provides a better platform for stacking the 16 images than the Fuji SW, but I don't know if the raw files produced can be processed. Perhaps bringing the files into C1 or LR, developing the raw files to taste (actually developing one and applying those adjustments to all 16 files), and then stacking in Helicon Focus may be the way to go.

I'm not sure about cloudy skies at slow shutter speeds. Just need to test. My gut says no, but can't think of a logical reason...
Any thoughts?
Helicon produces DNG which are highly editable, they conserve all the info you need. Id never edit raw first. always Helicon > Photoshop is my process

I have been using Helicon for 2 years and I can tell you...
I am the sloppiest photographer... i shot with the worst tripods at 1:1 magnification where even your breath moves the subject...
i live on the 5th floor near the subway stop, hardwood floor, old building...
once i even moved the focus manually on my Cambo Actus so imagine the damage in term so vibration...BUT
but the final results are magical...magical... everything is auto-aligned in a way that is unbelievable...

MUST be the sloppy FUJI software... they probably use only the averaging or what not
 
  • Like
Reactions: rga

gerald.d

Active member
I'm thinking that the pixel shift function CAN be useful for landscapes that do not have any movement (e.g., Cholla cactus in Joshua tree, Death Valley Canyons). Waterfall landscapes without foliage (Iceland?) may also work, though aliasing on the moving water would be interesting to test, esp. at slow shutter speeds. Same with a stationery object in water; slower shutter speeds.

Finally, I have read where Helicon Focus provides a better platform for stacking the 16 images than the Fuji SW, but I don't know if the raw files produced can be processed. Perhaps bringing the files into C1 or LR, developing the raw files to taste (actually developing one and applying those adjustments to all 16 files), and then stacking in Helicon Focus may be the way to go.

I'm not sure about cloudy skies at slow shutter speeds. Just need to test. My gut says no, but can't think of a logical reason...
Any thoughts?
Helicon Focus can be used to create a high resolution file from 16 pixel shifted images?
I had absolutely no idea it could do that. Would you mind providing a link to where you read that?

Kind regards,


Gerald.
 
OK, I tried today Helicon Focus

with Method A (average) not a bad result but worse then FUJI software
with Method C total garbage

so far, FUJI software wins.
so now I'm even more confused than before because this Pixelshift thing is totally unpredictable....

but all in all shortest interval and shortest speed 1/250 can provide good result in daylight
I dont have super fast strobes to prove it. my prodoto d4 heads are slow
 

Paul2660

Well-known member
Are you getting 400MP output from Helicon? I can see it aligning the images S in a focus stack but not giving the higher resolution as it’s not designed for that only stacking for focus. Just curious.

Paul C
 

rga

Member
So I did some testing today. I took the same image twice: once with pixel shift and one single shot. I processed the same pixel shift images using helicon and fuji software. The files were saved, after passing through ACR with no adjustments, as psd's:

Fuji Pixel Shift file size: 2.27GB [email protected]
Helicon Focus file size: 581.6MB [email protected]
Single Shot file size: 582.3GB [email protected]

Fuji Pixel Shift file enlarged to 81% was about the same size on my screen as the
Helicon Focus file enlarged to 149% as was the
Single Shot file enlarged to 149%

The images cropped without any alteration that I am aware of other than cropping the same small area and saved as .tif files
They are downloadable.

IMO as a non engineer/technical person the Helicon Focus image appears to have chosen the best pixels to create the file. The FujiPixelShift software has included every pixel and simply added them through a simple "algorithm" and the single shot is, well, a single shot. I would pick the HF file for large printing (40"x50" for me) as it seem to have a lot more definition already and would not take too much raw adjustment. If I were printing for Grand Central Station or larger, the FujiPS file could be manipulated easily enough to remove its softness, but still retain it's detail when resized. The single shot is perfectly fine if not a photo of still subjects...

Bob
 

f8orbust

Member
I'm confused. HF creates a final image from a vertical stack of images based primarily on the degree of focus in each image. With pixel shift you have a series of 'overlapping' images that all share the same focus point. I guess you can throw these images into HF, but I can't see anything it could offer in terms of processing since they are all equally in focus.

N.B. I've seen folks use pixel shift to take an image, change focus, use pixel shift again, change focus etc etc and then develop the individual image sets outside of HF before running them through HF to get a massive focus stacked image, but this is entirely different from using HF to 'develop' a single set of images.
 

Paul2660

Well-known member
What is being done, is it appears he is using Helicon to stack the images, then using an upsampling algorithm to up-rez the image. Helicon doesn't in anyway increase a output resolution. Based on his results, IMO it's not a bad idea to try as the output will have less noise which is true with any stacking process, and many modern up-rez software tools are quite impressive. Especially if you start with a very detailed sharp and clean image. The Fuji Software itself is a disappointment, at least for me. With the 1.5 years of time that Fuji has had to work this out, I feel they could have done better, but software solutions have always been a weak point for them, both with the GFX and X-series cameras. They have always relied on 3rd party developers to create the best raw conversion solutions.

Paul C
 

P. Chong

Well-known member
I think this is confusing. Helicon is a focus stacking tool. It DOES NOT give you a higher resolution image. If what is being done is just up rezzing, there is no need to take multi images which are shifted one pixel each. Just up-rez. This does not give you more detail. No matter how good the software is, it cannot create data to get more detail. Pixel shifting does get more data, and hence can give more detail.

Fuji has lots of time to get this right, but in my opinion, they did not. Hasselblad has pixel shifting technology as early as H4D-200MS in 2009, which makes one single raw image in camera. And Hasselblad is able to do it without resorting to electronic shutters, but with their leaf shutters, which mean that for use cases like table top and macro product photography, it works with studio strobes! So do many other brands - pixel shifted image in made in-camera via their electronic shutter. Like the Leica SL2 and Panasonic S1R which makes one regular resolution image plus one 187Mp high res image. Even Olympus with MFT does this in camera.


* edited to sort out grammar.
 
Last edited:

Ed Hurst

Well-known member
All this talk of 'finding ways to make it work properly' is rather disappointing, isn't it? When a 100MP camera has a feature like this added, one imagines that (if used in the right situations, such as static subjects) that it should 'just work' and simply deliver amazing 400MP results. Seems to me they have some work to do here... Any feature, however exciting, just might as well not be there if it can't be relied upon to work as advertised. Come on Fuji, make me annoyed that I don't have this camera!
 

gerald.d

Active member
I came to that conclusion too.
My setup with tripod and so on was probably fine. Using a shorter lens than the GF250 did not solve the issue alone. But I took the image on the 5th floor with wooden floors. After I went down to the basement I did better. But not perfect. So controlled environment seems to be essential. Difficult to do that in the field.

I attach the comparison at 200% and 1600%. The one from 5th floor and the GF250 on the left, the one from ground floor and the GF63 on the right. The lighting is different down there. So the image is less contrasty and sharp. But the artefacts are mostly gone.

View attachment 179229

View attachment 179230
OK let's try to bring this back on topic...

Thanks for your further testing Marc, but if I read correctly, there are two variables changing in this comparison - the lens and shooting location - yes?

I remain unconvinced that this is an external environment issue, and something has just occurred to me that I really should have thought of before.

The combing would appear to be far more dominant - in all examples I have seen - where the spatial frequency is high on the Y axis (i.e. the "comb teeth" appear to be vertical).

One way to test if this is rooted in an environmental problem rather than the camera system itself would be to shoot the same scene with the same equipment in the same environment with the camera in portrait and landscape orientations...

Kind regards,


Gerald.
 
Top