The Fujifilm Pixel Shift Multi-Shot Feature On The X-T5 And X-H2; How Good Are The 160mp Files?

This is one of the test images created from combining 20 exposures from my Fujifilm X-T5 using the
Pixel Shift Multi-Shot feature.  Unfortunately, you won't be able to see any of the fine detail,
sharpness or other attributes of this 15,456 X 10,304 pixel file.  The DNG file created is 698mb
in size!  Read below to find out what is good and bad about these files. (click to enlarge)

Since purchasing the Fujifilm X-T5, in the back of my mind I’ve wondered how well the Pixel Shift Multi-Shot feature works and how good the 160mp files really are?  I didn’t purchase the camera for that feature but I would be remiss if I didn’t at least try it out so if I ever wanted to use it, I would first, know how and second, understand what results I could expect.

For those of you who aren’t familiar with so called ‘pixel shift’ features, which are now available from several camera manufacturers, it uses the camera’s in-body-image-stabilization (IBIS) system to take multiple images in quick sequence and shifting the sensor about one pixel up, down, left, right, on a diagonal, etc. between each frame.  Then, either in-camera, or using special software for your computer, the image files can be combined into one very high resolution file.

I first encountered this feature on my Olympus E-M1 Mark III.  Olympus called this feature High Res Shot. One could create either an 80 megapixel, high resolution image when using a tripod or a 50 megapixel high resolution image using the camera handheld.  The Olympus camera takes 16 images and combined them into a final raw image. This is done in-camera. I thought it worked as designed but never really had a need for it.

Fujifilm has engineered theirs in a slightly different fashion than did Olympus. Fujifilm requires the camera be mounted on a tripod and the subject matter be absolutely free of any motion. That, in and of itself, very much restricts when and how often you can use this feature as well as on what subject matter.  Other camera manufacturers, who are on their second or third generation of this feature, have engineered their system to actually negate any minor motion that might be in the frame.  Also, as opposed to Olympus, Fujifilm cameras cannot combine the images in-camera, but requires you to download the Fujifilm Pixel Shift Combiner software to merge the image files after you return to your computer. Additionally, whereas Olympus takes 16 images, Fujifilm requires 20 images. I suspect that has to do with the X-Trans sensor and it’s nontraditional color filter array as opposed to the more common Bayer array in order for all the colors to be covered adequately.  Since this is Fujifilm’s first iteration of this feature, it is very basic.

How well does it work?  To find out I conducted a test to determine a) does it work, b) how easy is the feature to use, c) how do 26mp, 40mp and 160mp compare when it comes to fine detail and d), is it worthwhile to use.

 Well, it works as designed.  It is relatively straightforward to use if you have a sturdy tripod, little wind, a shorter focal length lens and absolutely no subject movement.  If there is any subject or camera movement detected, the software will 'fail' to combine the images.  That said, I found a serious issue in the 160mp images.  Also, I found the process, as compared to my former Olympus camera, cumbersome since Olympus does not require a tripod in all cases and combines the images in-camera as opposed to a separate piece of software.

Here is another example of a Pixel Shift 160mp file.  Again, the reduced size required by blogger
as well as the compression algorithms preclude you from seeing any significant detail in this file.
(click to enlarge)

I went out on a calm sunny day. With my medium sized Gitzo tripod, I took both my 40mp Fujifilm X-T5, which has the high resolution, Pixel Shift Multi-Shot feature and my 26mp Fuji film X-H2S, which does not.  I wished to compare identical 26 megapixel files to 40 megapixel files and also to 160mp megapixel files.  I set up everything as I normally would, cameras (alternatively) on the sturdy tripod, using the electronic shutter to eliminate shutter shake, a two-second delay to eliminate any potential movement from me pressing the shutter, aperture priority set at f/8 (previously tested to be the sharpest on the X-T5), raw file format and IBIS off.  I used two different lenses for the test, both of which are on Fujifilm’s list of recommended lenses for the X-T5.  Using the 16-55mm f/2.8, I made images at 35mm and 45mm, eliminating either the longest or shortest focal lengths to maximize potential quality.  On the 70-300mm f/4.5-5.6 lens, I used 300mm to test for camera/lens shake even on a calm day.  I ended up making five 160mp images, four with the 16-55mm lens and one with the 70-300mm lens.

(As an aside, remember last year when I had extensively written about the AF inaccuracy and inconsistency in the X-T4 as well as others reported in other X-series cameras?  Well, I have found none of that in either of these two cameras.  I have absolute confidence in the AF systems now as being fast, accurate and consistent.  Those old issues, evidently, have been rectified.  Thank you Fujifilm!)

Back at home, I transferred the images to a folder on my desktop.  I then downloaded and opened Fujifilm Pixel Shift Combiner software.  I looked on Fujifilm’s website to find instructions on how to best use the software but found none.  It only referred me to the help section of the software itself.  I found the software relatively easy to figure out.  I was able to set an output folder, which I set to be my “watched” folder for Lightroom.  The “watched” folder is one that LR monitors and anytime it finds an image in that folder, it automatically imports the images.  For each of the sets of 20 images I highlighted them, brought them into the Fujifilm software and hit the “run” command.  I then imported directly into LR the single 26mp and 40mp images for comparison to the 160mp images.

As I said, the software is pretty straightforward but the additional process required to combine the sets of images and then get them into LR is cumbersome, in my opinion.  Also, unfortunately, due to image size restrictions by Blogger as well as compression algorithms, you won't be able to see the subtle differences among the various regular size and Pixel Shift Multi-Shot sized files from the results of my tests, so I've tried to clearly articulate what I saw.

This is an extreme crop of one of the 160mp images created by Fujifilm's Pixel Shift feature.  Notice
the cross-hatch pattern?  It looks like a flannel shirt pattern! This pattern or artifacts can be found
 all over the final image.  Whatever Fujifilm engineered as far as the combining process, it needs fixed before the images can be used. (click to enlarge)

When looking closely at the 160mp images, I started to see some sort of artifacts (above).  I’ll include a highly cropped file in this post for you to see for yourself.  I’ll describe the pattern as sort of very fine vertical and horizontal lines.  Think what you would see if you magnified very fine cloth.  You would see vertical and horizontal interwoven strands.  At first I thought they were scan lines from the sensor until I looked closer and saw both horizontal and vertical, not just horizontal scan lines.

The other issue was that trying to sharpen the 160mp image files in LR just made the artifacts look worse.  I gave up.  I could not satisfactorily sharpen my test images in Lightroom Classic.  I then decided to try Topaz Sharpen AI and the result was much better.

After all of this, I decided to find out what quality and detail differences I would find if, instead of using the Pixel Shift Multi-Shot feature, I used Topaz Gigapixel AI and Topaz Photo AI to enlarge the image files to the same pixel dimensions as the Pixel Shift Multi-Shot files.  I closely examined the files and chose the best settings in each of those programs and after very close evaluation of those as compared to the Pixel Shift Multi-Shot files, the Topaz files actually looked better.  Also, I was not able to discern any additional fine detail in the Pixel Shift Multi-Shot files that wasn’t in the other Topaz upsized files.  The bottom line is the Topaz files looked better than the Pixel Shift Multi-Shot files.  They were sharper, had no artifacts and had the exact same detail, in my opinion.

Finally, I wanted to just look at the detail differences between the 26mp files, the 40mp files and the 160mp Pixel Shift files.  As I found when comparing identical images using my 47mp Nikon Z7II to the 26mp X-H2S files, if you look really, really closely you can find a few differences in detail when comparing the X-H2S to the X-T5 files.  Finding more detail in files, all technical aspects aside, is entirely dependent on the subjects in your photographs.  In this case it was like the game, “Where’s Waldo.”  I really had to search to find any differences in fine detail.  I did find some, but as I said, much of it is subject dependent.  Conversely, it might be that the detail in the X-T5 when rendered at 100% was just larger than the detail rendered at 100% in the X-H2S files rather than seeing actual more detail; it is easier to see because it is larger.  Having to look that closely to see any differences tells me that in 99.9% of the cases, it will make no practical difference at all for most of you.  Save your money if you think there are massive differences in detail.  Buy the X-T5 for other reasons instead.  There are a myriad of other good reasons to buy this camera.

This image is cropped to 1.5% of the original 160mp file.  So this is pixel peeping at its 
extreme!  But I'm showing it to you to make a point.  This image is the original 40mp
file from the X-T5 upsized in Topaz Photo AI to 160mp. (click to enlarge)

This image is the same file but using the Fujifilm Pixel Shift Multi-Shot feature, has been
upsized to 160mp in Pixel Shift Combiner software.  In my looking at the two files on my
27" monitor, the Topaz file is sharper and has just as much fine detail as the 160mp file.
(click to enlarge)

As far as the 40mp files versus the 160mp files, well, those artifacts make those 160mp files useless in my opinion.  Also, I had to work really hard to successfully sharpen them so fine detail is much harder to extract.

The bottom line for me is that I cannot recommend using the Pixel Shift Multi-Shot feature because of the defects and artifacts that are found in the combined image files, the extra many steps in workflow that are required to combine the 20 images in each set as well as the very limiting parameters set for its use.  One image upsized does not have the limitation of using a tripod and having your subject absolutely motionless.

My current plans are to continue to use both Topaz Photo AI and Topaz Gigapixel AI if I need a file size upsized. That also adds the versatility of other file sizes besides only 160mp.  (I’m not sure if I will ever actually need such a large file.)  I’m hoping Fujifilm will go back and re-engineer this feature so it can match other camera manufacturers methodologies as well as remove the artifacts from the combined image files.

DISCLAIMER:  I have no affiliation, formal or informal, with Fujifilm or any of the software manufacturers I have mentioned here.  I have no affiliate links, no click-through sites and am not paid in any way.  I buy all my gear and software just as do you.  I have 52 years of photographic experience and just share what I discover and what I know hoping it will help others enjoy photography as much as I have.  Also, all of this is my opinion formed from my tests with my gear and I would stress you conduct your own tests and generate results for your cameras, lenses, software and your techniques.

Join me over at my website, https://www.dennismook.com.  

Thanks for looking. Enjoy!  

Dennis A. Mook  

All content on this blog is © 2013-2023 Dennis A. Mook. All Rights Reserved. Feel free to point to this blog from your website with full attribution. Permission may be granted for commercial use. Please contact Mr. Mook to discuss permission to reproduce the blog posts and/or images.