Pixel Shift: A Primer for Beginners
Feb 19, 2020

Having answered several PM’s over the years on the subject of pixel shift, I thought I would combine the common questions and post a thread on how I use pixel shift. There are many…let’s call them opinions, on what can and cannot be done with pixel shift. Let’s use this thread to discuss the practical uses and limitations.

For background, prior to pixel shift, I carried an A7R and later A7RII with compact primes alongside mFT for scenes where I desired more resolution and dynamic range. That practice eventually changed in 2016 as I gained experience with pixel shift on the PEN-F and later, the EM1mkII. Let’s be clear, it will NOT turn your PEN-F into a D850, but under the right circumstances it does produce far better image quality than you might expect from a crop sensor camera.

Dealing With Motion:

Let’s put the obvious issue at the forefront. Yes, you can have motion in your scene, but not all motion will compliment pixel shift. You might have seen the artifacts that appear when movement occurs. How do I avoid those? Let’s start with the simplest scenario. When shooting a still scene as below, I use whatever shutter speed comes with the best aperture and lowest ISO while keeping the camera motionless. Easy enough!

Static scene - wanted additional DR to see underside of roadway without blowing out the sky.

In scenes with movement however, I tend to live at the extremes of shutter speed. Either as fast as possible, or so slow, it doesn’t matter. With large aperture (shallow depth of field) I usually end up pushing shutter speed (and sometimes ISO) higher in order to complete the 8-shot sequence before movement in the scene occurs. When I need to maximize depth of field, I often resort to long exposures. I find this coincides nicely with light levels at the times I prefer to shoot (dusk and dawn). It certainly doesn’t preclude one from using “in-between” shutter speeds for say landscapes in broad daylight at 1/200th, it’s just that you increase your likelihood for motion artifacts if there’s movement. These portions of an image can be masked out by overlaying the standard resolution ORI file (rename to ORF) if need be.

Long exposure to smooth out the waves

Environmental Conditions I Avoid:

I avoid pixel shift in unpredictable, gusting winds, pedestrian/auto bridges and shooting from anything suspended in or above water, like a dock. This is one area hand-held high-res is really handy but beyond the scope of this thread. Ultimately, the camera has to be motionless for this to work so avoid working on anything that moves beneath you. Sand can be tricky – I avoid small contact area on soft sand. Either bury the tripod feet as deep as you can or find a wide stable base (driftwood, surfboard, camera bag etc) to set the camera on.

How Do Results Compare?

Quite favorably. You’ve probably seen the difference between standard 20MP and 50 or 80MP high-res (examples below). More detail, cleaner shadows, etc.

A 1:1 example of the additional image quality present in pixel shift (left) vs standard shot (right @200%)

What about high-resolution full frame cameras? If you’ve edited full frame RAW files you will appreciate the additional latitude pixel shift images afford. Want to lift the shadows 5-stops? Go for it!

Shot at equivalence but underexposed 5EV then lightened in post. The pixel shift image on the left has effectively one additional stop of light.

If I had to draw similarities, I would say the output of pixel shift typically resembles a 40-something MP FF image (think A7RIII, Z7, S1R). This is probably why Olympus caps the JPEG output at 50MP. I’m not convinced you’re getting usable resolution beyond 40-50MP in most cases.

1:1 view - high-res on the right downsized to match A7R2. Overall similar, maybe a little more detail in the pixel shift image.

Perhaps with a very good lens, there is a little more detail in it (Nocticron at f/4):

Viewed 100%. When downscaled to 60MP or so, the pixel shift image looks very similar to A7RIV with regard to detail.

What Lenses Do I Use?

Whatever you have with you! Don’t be afraid to experiment. As with any camera/lens combination you want to extract the most detail from, it helps to know the apertures your lens performs best. Often the subject will determine your aperture for you but don’t be afraid to use the lens you have with you. Even a relative mediocre performer like the Olympus 17/1.8 will benefit from pixel shift vs standard res; the 17/1.2 even moreso.

1:1 view. Looking at the numbers printed around the BASS knob, there is obvious benefit from pixel shift (left vs center). The 17/1.2 at the same aperture (right) is better still. Note: the 17/1.2 was shot from the exact same position as the 17/1.8 but the effective magnification is rather different between these lenses for near subjects.

Another example from an adapted Pentax 24/2.8 worth about $10 vs the Olympus 25/1.2.

1:1 view. Even an old adapted lens benefits from pixel shift, but there is more to be gained from using a sharper lens like the 25/1.2

Smaller primes are often easier to balance and stabilize but I have experienced excellent results from larger zooms as well.

1:1 portion of a photo from ~ 350 meters away on an overcast day. Shot with the 40-150/2.8.

1:1 view sandbox toy, shot with the 12-100. Yes, it's over-exposed (sand in direct sunlight).

What Do I Use It For?

I use it primarily landscape, seascape, architecture, urban scenes, still life, product photography (example gallery here). Occasionally macro, archiving, real estate, and light trails. Perhaps my favorite use for pixel shift is long exposure scenes. It’s a shame that Panasonic limited the G9 to 1s. The early Olympus cameras like the EM5mkII and PEN-F were limited to 8s which is usually sufficient for what I shoot. The later OM-D’s go all the way to 60s.

Once you get into long shutter speeds (1s and longer) the resulting image looks similar to the effect of a ND filter. Not identical, but similar (top right vs bottom left picture below). You can take it one step further and combine pixel shift with an ND filter to really exaggerate the effect (bottom right).

Shot with the 12-100 with and without 6-stop ND filter

What Tripod Do I Need?

Let’s put the “must have sturdy tripod” mantra in context. A solid, heavy tripod is no doubt beneficial to gain the most from any camera but that’s not always conducive to life. Use what you have, but I wouldn’t further encumber yourself with a $1000 Gitzo just because you might try pixel shift on your next outing.

I travel a lot for work and leisure, forgoing a tripod for mobility. When I do carry a tripod it is a pocketable, plastic UltraPod. Otherwise I use what the environment affords: rocks, bricks, logs, camera bag, paint cans, ledges, you get the idea. How big of a difference does it make?

Just to to make some readers cringe - an empty aluminum can for pixel shift!

1:1 view - taken with the PEN-F and 45/1.8. One of these was shot with my largest, heaviest tripod. One with a $10 UltraPod II. One resting on a camera backpack. One sitting on an empty aluminum soda can Can you guess which is which?

To be clear, I would avoid something as vibration prone as an aluminum can but I was curious what the results would look like (it's A if you couldn't tell. Look at the jaggies around the 1/40s marking ;-))

Why Do RAW Shots Look Soft?

The 50MP JPG’s produced in camera are excellent but if you want to further play with the RAW, a word of caution; while Adobe Camera RAW will open a high-res ORF file, it’s probably not where I would start. The DPR Studio Scene “simulated RAW’s” are a great example of this. To say Olympus Workspace (OW) is better at retaining detail is a gross understatement – see for yourself. There is no amount of magic I can work in Adobe to make the pixel shift Olympus images look like they do straight out of OW.

1:1 view of a wine label. Same file, but converted differently. Taken with an EM1mkII and 45/1.8

1:1 view of a cathedral window. Same file, but converted differently. Taken with the EM1mkII and 12-100/4.

1:1 view from the studio scene. Same file, but converted differently.

How Should I Sharpen for Final Output?

Entire threads have been devoted to optimal sharpening of pixel shift images. Everyone has their workflows and preferred methods so I can’t say what is right or wrong for you. I will advise you to start your RAW workflow by converting the ORF into TIF via Olympus Workspace first, in order to save yourself some headaches later. Pixel shift images require more aggressive sharpening than you might expect. Every photograph is different. I have a preset saved as a starting point in Lightroom: Amount: 75, Radius: 2.7, Detail: 3, Masking: 50 and will adjust accordingly from there. I have seen great results from unsharp mask in Photoshop and the AI Module from Topaz as well. This part takes some experimenting to find what works for you.

Final Words:

There are many creative uses for pixel shift beyond what I first anticipated. Like many, I had my doubts as to the utility in the field. What I’ve learned is that like any camera feature, there are circumstances that will benefit and those that are best avoided. My advice is to experiment with the suggestions above and see for yourself if it’s something you will benefit from. Feel free to post questions, suggestions and tips below. I do ask that we try and keep commentary constructive and on topic. Thank you. -IM