Pixel-Shift Shootout: Olympus vs. Pentax vs. Sony vs. Panasonic

15Share

Over the past few years, pixel-shift image capture has transformed from a luxury reserved for deep-pocketed specialists to an increasingly common feature on new, resolution-oriented cameras. Today, in addition to Hasselblad’s behemoth Hasselblad H6D-400c, Olympus, Pentax, Sony, and Panasonic offer versions of the technology at much more accessible prices.

While it is advertised under several names (High Resolution Mode, Pixel Shifting Resolution System, Pixel Shift Multi Shooting Mode), the basic action of pixel-shift image capture remains the same: move the camera’s sensor and capture more data. I set out to test the power (and limitations) of the versions of this technology offered by the Olympus OM-D E-M1 Mark II Mirrorless Micro Four Thirds Digital Camera, Pentax K-1 DSLR Camera (just replaced by the Pentax K-1 Mark II), Sony a7R III Mirrorless Digital Camera, and Panasonic Lumix DC-G9 Mirrorless Micro Four Thirds Digital Camera.

How does it work?

Gauging the usefulness of pixel-shift image capture for your photography first requires an understanding of how it works—and how digital sensors, in general, work. This topic could span many articles, but I will keep it (relatively) short and sweet. Where analog cameras use light-sensitive film to record images, digital cameras use light-sensitive photodiodes. In most digital cameras, each photodiode is paired with a specific color filter (red, green, or blue), forming a photosite. The arrangement of red, green, and blue photosites on a sensor follows a specific pattern known as a Bayer array (a.k.a. Bayer matrix, filter). Twice as many green photosites are used as red or blue because human vision is most attuned to resolving detail in green.

Figure 1: Bayer array pattern

Because each photosite is sensitive to only one color, the data that it records is incomplete. To solve this problem, the collected data is interpolated via a de-mosaicing algorithm either in-camera or on a computer, a process that assigns values for all three colors for each photosite based upon the collective values registered by neighboring photosites. The resulting colors are then output as a grid of pixels and a digital photograph is born.

NOTE: Photosites are often confusingly (and incorrectly) described as pixels. Even the term “pixel-shift” is misleading. Pixels do not move; sensors move. There are no pixels on your camera’s sensor—only on the images that it produces. Photosites are physical locations on a camera’s sensor, pixels are abstractions of data used to form digital images.

Figure 2. In single-image capture, each photosite records data for red, green, or blue light. This data is interpolated by a computer so that each pixel in the resulting digital photograph has a value for all three colors.

Shifting Pixels

Pixel-shift cameras attempt to reduce the reliance on interpolation by capturing color data for red, green, and blue for each resulting pixel by physically moving the camera’s sensor. Consider a 2-pixel by 2-pixel square taken from a digital photograph. Conventional digital capture using a Bayer array will record data from four photosites: two green, one blue, and one red. The missing color values for each site will be determined during interpolation. Now consider a 2-pixel by 2-pixel square on a digital photograph created using pixel-shift technology. The process begins as described above; data is recorded from the four photosites. Next, the sensor “shifts” in 1-photosite increments three times. During this process, data from four photosites (two green, one red, one blue) is acquired for each pixel, resulting in more robust color values and less intrusive interpolation.

Figure 3. Pixel-shift technology works by physically moving the sensor so that each pixel in an image is created by combining data from four (or in some cases, eight) photosites.

Sony’s Pixel Shift Multi Shooting Mode and Pentax’s Pixel Shifting Resolution System operate as described above. It is important to note that using these modes does not increase the total number of pixels in your final image. The dimensions of your resulting files remain the same, but color accuracy and detail are improved.

The High Resolution Mode of Panasonic and Olympus cameras, which both use Micro Four Thirds sensors, takes a slightly more nuanced approach, combining eight exposures taken ½ pixel apart from one another. Unlike Sony and Pentax, this doubles the number of pixels in the resulting image.

Limitations

The greatest challenge facing pixel-shift image capture is motion. The process requires, at minimum, four times the exposure time of single image capture. This translates into four opportunities for a part of your composition and/or your camera to move during image capture, degrading image quality. Such constraints limit the technology’s application to still life and (static) landscape photography. Additionally, trying to pair a strobe with a camera using pixel-shift image capture can be complicated by the speed of image capture, flash recycle limitations, and general compatibility problems. Manufacturers are aware of these problems, and, as will become apparent in the comparison below, are working to resolve them.

Comparison Conditions

For my test, I wanted to see not only how the pixel-shift image capture of each camera compared to its standard image capture, but also how the results shook out between brands. Each camera was paired with its proprietary 35mm equivalent, 16-35mm zoom (see chart below), a proprietary shutter release cable, and mounted on the 3 Legged Thing Equinox Albert Carbon Fiber Travel Tripod with AirHed 360 Ball Head. Despite best intentions, the resulting images exhibit some variation in framing and focal length worth keeping in mind when looking at the camera-to-camera comparisons below. Expressed as 35mm-equivalent values where appropriate, they are: Olympus (28mm), Pentax (30mm), Sony (35mm), and Panasonic (36mm).  Otherwise, shooting conditions across cameras were held constant: shutter speed, aperture, and ISO were all held constant. White balance was set to AUTO, which led to a bit of variation in the colors rendered between brands—but that is the topic of another article. Each starting image is JPEG out-of-camera—or, in the case of Sony, out-of-software. Note: the a7R III does not produce out-of-camera pixel-shift images; you must run the four raw files through its (free) proprietary software, Imaging Edge. I made no further adjustments in the software.

Camera Lens Single Capture Image Size Largest Image Size Possible Output Resolution pixels/inch Number of Exposures
Pentax K-1 Pentax HD FA 15-30mm f/2.8 ED SDM WR Lens 7360 x 4912 7360 x 4912 300 4
Sony a7R III  Sony FE 16-35mm f/2.8 GM Lens 7952 x 5304 7952 x 5304 300 4
Olympus OM-D E-M1 Mark II  Olympus M.Zuiko Digital ED 7-14mm f/2.8 PRO Lens 3888 x 5184 7776 x 10368 180 8
Panasonic Lumix DC-G9  Panasonic Leica DG Vario-Elmarit 8-18mm f/2.8-4 ASPH. Lens 3888 x 5184 7776 x 10368 180 8

Comparison with Single-Image Capture

For this part of the test, I shot the Manhattan Bridge, a complex, but relatively static subject. The good news for photographers who do not have a camera with pixel-shift capability is that it takes a bit of searching to find noticeable, qualitative differences between single-image and pixel-shift capture images—a testament to the accuracy of current interpolation algorithms.

Figure 4. Olympus OM-D E-M1 Mark II complete frame, single image (left) and pixel-shift (right)
Figure 5. Olympus OM-D E-M1 Mark II detail, single image (left) and pixel-shift (right). Note that this comparison reveals the discrepancy in size between single-image and pixel-shift images created by this camera.
Figure 6. Pentax K-1 complete frame, single image (left) and pixel-shift (right)
Figure 7. Pentax K-1 detail, single image (left) and pixel-shift (right)
Figure 8. Sony a7R III complete frame, single image (left) and pixel-shift (right)
Figure 9. Sony a7R III detail, single image (left) and pixel-shift (right)
Figure 10. Panasonic Lumix DC-G9 complete frame, single image (left) and pixel-shift (right)
Figure 11. Panasonic Lumix DC-G9 detail, single image (left) and pixel-shift (right). Note that this comparison reveals the discrepancy in size between single-image and pixel-shift images created by this camera.

Pixel-Shift Image Comparison Between Brands

To make the difference in image quality apparent on-screen, I pulled the same small area out of each image. Since each sensor produces images of a different size, the first comparison below shows each image in its “native” size out of camera. The second comparison shows the result of transforming each detail shot so that it matches the largest output image (Panasonic). The third comparison shows the result of transforming each detail shot to match the smallest output image (Olympus).

Figure 12. Detail comparison at output size from each camera
Figure 13. Detail comparison, all files transformed up to match size of Panasonic
Figure 14. Detail comparison, all files transformed down to match size of Olympus

Motion

Movement poses a significant obstacle for pixel-shift technologies. Of the cameras tested, only the Pentax K-1 offers a “Motion-Correction” mode when working in the setting. This is an in-camera fix; raw files taken in this mode reveal Bayer distortions, while JPEG files relay a moving object only in the first frame in the sequence, eliminating it from the rest.

Figures (15 & 16). Comparison of motion effect on Pentax K-1 raw capture converted to JPEG in Photoshop (left) and in-camera JPEG capture (right)

Motion had a noticeable effect on image quality for the other brands tested, significantly limiting their potential applications.

Figure 17. Movement in frame across brands

Artificial Light

I wanted to see if any of the cameras could create a pixel-shift image while using strobe lighting. When paired with a Pocketwizard Plus III Transceiver, all of the cameras were able to trigger a Profoto D1 Air 500 W/s Monolight for single image capture, but only the a7R III was able to trigger the light in pixel-shift mode. Despite triggering the light, the syncing was off, resulting in only partially lit, unusable images. However, when I attached Sony’s FA-WRC1M Wireless Radio Commander, paired with the FA-WRR1 Wireless Radio Receiver, triggering was seamless and I was able to get four equally lit images. Proprietary shenanigans aside, this is possible because the a7R III pauses between each image capture, whereas the other cameras tested are nearly immediate.

Camera Pixel Shift Image Out-of-Camera? Compatible with Radio-Triggered Lights? Access to Individual Files? Motion Correction?
Pentax K-1 DSLR  Yes No No Yes
Sony a7R III  No Yes Yes No
Olympus OM-D E-M1 Mark II  Yes No No No
Panasonic Lumix DC-G9  Yes No No No

Final Thoughts

The most telling revelation from this experience was less the differences in image quality between brands and more the impressive job that each camera did while operating in single-image capture. While the images made using pixel-shift capture were of a greater quality, the limitations of the technology keep its applications rather narrow.

One area where improvements could be made across brands is in user control. While the requirement of downloading extra software to create pixel-shift images with Sony is a nuisance, Imaging Edge does allow you to get a little more hands-on by adjusting peripheral edge noise settings when creating your image. Expanding user-input (e.g. managing areas of motion, time between exposures) would be a huge help for photographers looking to get the most out of the technology.

What do you think of pixel-shift image capture? Have you used it? Do you love it? Do you hate it? Let us know in the Comments section, below.

15 Comments

Julien P. Has made an important point. Really can not compare images for 'resolution' if they were taken at different focal lengths let alone different glass. Just what are we comparing? Some basic flaws here in methodology.

"Figure 12. Detail comparison at output size from each camera" looks wrong.
For example, K-1 outputs 7360 x 4912 whereas A7R III outputs 7952 x 5304 .
That means the full Sony image should be about 8% larger in each dimension vs Pentax. That is the difference in sensor resolution.
I could tell with the naked eye that the Sony image looked too large relative to the Pentax.

So, I downloaded the figure 12 image and did crops.
A crop of the K-1 image done on my PC reveals it is about 224x188 .
Whereas the Sony image is 295x247.

The Sony image should be 8% larger in each dimension. But it is actually about 31% larger.
This probably comes down to different focal lengths or lenses being used.
The K-1 image was shot with a 15-30 zoom, and A7R III with 16-35 zoom. B&H should have been able to match the focal length between those two cameras, but they didn't.
The Sony was zoomed in much closer than the Pentax. Of course it will have more detail, in addition to having a higher resolution sensor.

Hi Julien,

That is a fair point. The Pentax image was shot with a wider lens so when you look at the original images, you will see more in the frame. There were more variables (and unavoidable compromises) at work behind the scenes during this test than I would have preferred. One of them involved selecting lenses to use of comparable specs so as to avoid overly influencing the sharpness of the images. In an ideal world, all of the cameras would have had the same sensor size, allowing me to adapt a sharp prime to use across the board and remove that variable. Ultimately the discrepancy that you noticed was a result of human error on my part from trying to frame the same shot with a bunch of zoom lenses while outracing a rainstorm with all of that gear. Thanks for reading-- and your sharp eyes.

Cory,

Would you mind telling us what focal lens was shot with each camera ?

The Sony A7R III and Pentax K-1 have the same sensor size. Your tests used 15-30 and 16-35 lenses. You could have kept the focal length identical between those two cameras by using anything in between 16-30mm . For the micro 4/3 cameras, you would have to use something between 8-15mm to match the FF sensor cameras.

The micro 4/3 lenses were 7-14 and 8-18 for these cameras. So, anything between 8-14mm on the micro 4/3 would have worked to match all 4 cameras to equivalent focal lengths. Of course, you would have to compare parts of the images in the center, since the micro 4/3 cameras crop the left & right sides compared to the FF cameras due to different sensor aspect ratio.

Would you mind adding the focal length you shot for each camera to the review ?

Hi Julien,

The text has been amended to include the variations in focal length between images for reader reference. Thanks again for pointing this out. 

Great article. I've always wondered how the Olympus Hi Res mode stacked up to the others on the market. I use it when shoot studio scenes and since updating to the latest firmware, the EM1 MkII works perfectly with the GODOX XT1 wireless trigger. One thing to note if you are using it with strobes is the systems aperture limits. The Olympus system limits you to no more than f8. So if you are shooting non Hi Res images at f11 and then switch to Hi Res mode, you will need to adjust your flash output.

As an Olympus fan I am quite impressed with the native resolution, apparent sharpness, and details in shadow areas on Sony a7R3.  As Olympus prices go up the value of the buck spent for A7R3 increases.  However, searching for birds, nature subjects, optical viewfinders are valuable, more successful in following a moving subject.  Pentax, in this test, the only candidate with OVF— but disappointing sharpness, with or without sensor-shift.  Looks like more than one body system needed to exploit digital photography completely!

good read

The HRS does work with studio strobes.  You need to use the PC connector.  The EM5mkII worked with strobes and the EM1mkII worked with strobes after the update.  The RAW file on the EM1mkII is 10,368x7776 pixels.  When using HRS with PC connector the shutter speed needs to be 1/13th of a second.  Use is regularly.

Hi Jay,

Thanks for sharing this info. I'm glad to hear that there is a workaround for strobes. Hopefully, as the technology advances, it will allow more flexibility from its users. Also, thank you for catching the RAW file info. The table has been updated accordingly. 

You also forgot to mention that the Pentax K-3II, Pentax XP and Pentax K-70 support pixel shift.

Hi Philip,

Pentax is definitely committed to the technology. As fate would have it, almost immediately after the shooting for this article was completed, the K1-II, which has Pentax's latest version of the technology, was announced. Thanks for reading!

Excellent basic explanation of the technology, but it falls a little short on details. The E-M1 Mk 2 produces the same size file as the Panasonic, and you neglect to mention that there is significant noise reduction and improvement in dynamic range using sensor shifting.

Hi William,

Good catch on the Olympus; the table has been corrected. As for noise reduction and dynamic range-- those are definitely important benefits of the technology that I hope the test images illustrate. Thanks for reading!

Your test images have not been updated to reflect the larger size yet, though. You still use the smaller Olympus file, which makes it look terrible compared to the others.

Close

Close

Close