Olympus E-M5 II High-Resolution Mode

Exploring the E-M5 II's "High Resolution" mode

by Dave Etchells and Dave Pardue

For photographers interested in shooting high resolution imagery for landscapes, architecture or studio product photography, the E-M5 II comes equipped with a special high resolution mode designed to yield a 40 megapixel JPEG image competitive with those available from 36MP full-frame cameras like the Nikon D810 and Sony A7R, and a 64-megapixel RAW file. Really? A camera with a 16-megapixel sensor is going to somehow output 40/64-megapixel files? From our first briefing on this mode, we've been impatiently waiting to put it to the test in our lab! After all, the Sony A7R was our overall Camera of the Year for 2013, and the Nikon D810 won that honor just last year, so this is a rather bold claim for a Micro Four Thirds camera to make!

Before looking at the results (OK -- go peek if you'd like and then come back here) let's take a quick glance at the mechanics of this feat. It's actually pretty clever; one of those things you wonder why it hadn't been done already, once you learn what they're doing. First and foremost, be advised that you must use a tripod in order to take advantage of this new mode, and your subject(s) must be motionless unless you're deliberately aiming for odd motion blur effects.

In High Resolution mode, the E-M5 II fires eight sequential images across a roughly one-second time period using the electronic shutter, while at the same time using its voice coil-powered sensor-shift image stabilization system to move the sensor very slightly between shots. The net result is eight images, shifted by sub-pixel amounts from each other, that are then combined into the final super-res image. The end result is either a 40-megapixel JPEG file or that JPEG plus a 64-megapixel ORF file and a 16MP ORI file (not yet supported in common post-processing software). The 64MP ORF file can only be processed with a special Photoshop plug-in, which we haven't seen yet as of this writing.

As you'll see below, one positive side effect of the E-M5 II's "oversampling" of the image is that it pretty much completely eliminates moiré. The multiple exposures also significantly reduce image noise at the ISO 1600 limit imposed in this mode. The overall results are flat-out amazing, subject only to the limitation of completely stationary camera and subject.

OK, let's get to the fun part: the comparisons!

[Editor's note: all crops are from in-camera JPEGs with default settings. The E-M5 II images have been slightly down-sampled in order to show a similar size for the comparisons. Clicking on any image will take you to the straight-from-the-camera lab test image. You may also access the RAW files for most of these from each camera's lab samples page.]

Update: Based on a helpful reader request, below are true 1:1 crops to show the E-M5 II at the full 40mp alongside the D810 at 36.3mp:

In high-resolution mode, the E-M5 limits your ISO choices to a maximum of 1600. Let's take a look at that; the results are again surprising.

The conclusion from our lab testing is that the Olympus E-M5 II's new high resolution shot mode is truly ground-breaking for applications involving shooting non-moving imagery from a tripod. Landscape photographers, architectural photographers and studio product photographers can rejoice! It's flat-out amazing to see this kind of performance from such a compact, affordable camera!

Exploring the E-M5 Mark II's High-Res Shot Mode RAW files

by William Brawley and Dave Etchells

The High-Res Shot mode on the new Olympus E-M5 II offers not only 40-megapixel JPEGs (39.9MP, 7296 x 5472 pixels), but also the option of very high-resolution 64-megapixel RAW files (63.7MP or 9216 x 6912 pixels).  Olympus provides a free High-Res Shot plug-in for 64-bit Adobe Photoshop CS5 and newer (note that we were originally told CS4 and newer), to allow for editing to your heart's content. Below are some conversions we've made with it as well as some initial impressions using it.

In use, for all intents and purposes, the Olympus PS plug-in takes the place of the Adobe Camera Raw intermediary step that you'd see when opening a RAW file in Photoshop, and provides a similar, though vastly limited set of adjustments options such as exposure compensation, white balance, contrast, sharpness, and saturation. The plug-in also allows for both single image and batch file processing. Once processed, the RAW file can be opened and further adjusted in Photoshop just like any other converted RAW image file.

Now for our side-by-side comparisons. While our previous comparisons of JPEGs compared to the Nikon D810, its 36-megapixel resolution images are noticeably different to the E-M5 II's 64-megapixel high-res RAW images. Here, we instead opted to compare to the Pentax 645Z, which produces 51.1-megapixel images, the highest native-resolution camera we've tested thus far. In the 100% crops below, one set of crops are displayed with minimum sharpening and no additional post-processing performed to give you an idea of the relative sharpness between the two cameras, while the second pair shows images with some unsharp mask sharpening applied in Photoshop.

(Note: Click the Olympus image crops below to download the full-size converted JPEG files. We're also providing links to their respective RAW files. As of this posting, the Olympus High Res Shot Raw Photoshop Plug-in is available to download, but one must input a valid serial number for an E-M5 II in order to initiate the download. The Pentax 645Z crops below link to their corresponding RAW files.)

In High-Res Shot mode, the E-M5 II limits your ISO choices to a maximum of 1600. Let's take a look at that again, this time with RAW files.

The conclusion from our testing with High-Res RAW files is that although the Olympus E-M5 II's new High Resolution Shot mode doesn't produce files with as much detail as some (larger and much more expensive) cameras with lower native resolutions like the Pentax 645Z, it is still truly ground-breaking for applications involving shooting non-moving imagery from a tripod. Landscape photographers, architectural photographers and studio product photographers can rejoice! It's flat-out amazing to see this kind of performance from such a compact, affordable camera. Stay tuned, as we'll take it out into the real world and run it through its paces there as well.

How'd they do that? A little technical aside from Dave Etchells:

Right off the bat, I'm sure everyone is going to be asking how you can in any meaningful sense get a 40-megapixel image out of a 16-megapixel sensor. After all, the image sensor is looking at the world as a grid of 16 million little windows, and any subject detail within each of those windows is averaged together. The image comparisons above are pretty telling, but how can they tease out that kind of detail, if the sensor only has 16 megapixels to begin with? I mean, the subject detail within a given pixel could be half white and half black, or the whole area a medium gray, and the sensor would have no way of knowing the difference.

Well, the camera has no way of knowing what's going on at a sub-pixel level, unless it moves the sensor around a bit and watches what happens. Imagine the case just mentioned, where we have a sharp black/white boundary in the subject (let's make it a vertical edge, for the sake of illustration) which happens to be exactly straddled by a sensor pixel. As noted, that pixel is going to see a middle gray. But what happens if we shifted the sensor a half-pixel left and right? Shift it one way, and the pixel will be all the way on the black side, so it will "see" black. Shift it a half-pixel the other way, and it'll be all the way on the white side, so it will "see" white. Knowing that the three images were grabbed with a half-pixel shift in between, the camera can infer that there must have been an abrupt black/white transition in the subject, centered on the location of the original pixel.

The case of a simple black/white transition is easy to figure out, but what happens when there's all sorts of tonal and color detail changing across the image? How can you sort all that out? It turns out there's some fancy math to do just this; the technique is called deconvolution. It's been in common use for several years now, originally developed for use in reflection seismology, but used extensively in many forms of imaging. NASA in particular has used it to sharpen and correct images from planetary spacecraft and even the Hubble Space Telescope (before it got its corrective "goggles" installed). It's also been used in microscope imaging, and in fact by Olympus themselves. Here's an example from Olympus' own microscopy site, showing the original image on the left, the processed one on the right. It's pretty amazing how much fine detail can be extracted from a fuzzy image!

So that explains what's likely going on inside the Olympus OM-D E-M5 II. As someone who was involved with image processing as far back as 1977 or so, it's amazing that this capability can now be built into a compact, portable camera. (The E-M5 II almost certainly has way more processing power than NASA "supercomputers" from the '70s!)