A load of old pixel shift. Why I just don’t care for high-res modes: Digital Photography Review

This is probably going to be an unpopular opinion, but I’d argue that multi-shot high-resolution modes are all but pointless.

Not entirely pointless: I’m sure there are applications out there where the subject stays still enough and where it’s practical to bring a tripod, and the improvement is worth the effort. And if you’re someone whose photography it suits, then I’m happy for you. But that’s very rarely the case, so it’s just not a feature I find very useful, and it’s certainly not one I’d use to choose one camera over another.

A lot of this is because most implementations are terrible. Noticeably, every brand appears to have a slightly different implementation, with the exact approach, the number of shots and whether they can be combined in-camera differing. This strongly points to it being an area in which each manufacturer is patenting its own approach and blocking others using it, and in doing so, making it harder for anyone else to deliver a usable/useful version.

Lost in a thicket of patents

It’s no surprise that some of the earliest pixel shift modes came from Olympus and Pentax: two companies that committed early to the idea of in-body stabilization, both looking for other features it could be used to deliver.

The Pentax system shoots four images, canceling out the Bayer pattern so that it has full-color information for each output pixel. This delivers greater color resolution with less aliasing, greater sharpness (through lack of demosaicing) and the improvement in noise that comes from combining multiple images. These individual shots are combined into a finished output file.

The Olympus method shoots eight images: four canceling out the Bayer pattern, then moving to position 1/2 a pixel offset from this and repeating the process to quadruple the output resolution. This system also combines the images in-camera and is unique in letting you set a delay if you were shooting, say, product photos and needed time for your strobes to recycle between shots.

Panasonic currently has one of the best (least-terrible?) high-res implementations. There’s an eight-shot mode with the option of 4X or 2X the normal pixel count, with or without motion correction, or a handheld mode. But the fact that it’s essentially three modes, each with its own trade-offs, hints at how far from ideal they all are. At least they’re combined in-camera, though. Panasonic G9 II | Panasonic 12-60mm F2.8-4 Asph OIS | ISO 100 | F5.6 | 1/320 sec Photo: Jeff Keller

It’s been mostly downhill from here, though. Sony first adopted a Pentax-like four-shot mode that had to be combined on a computer, before later adding a more Olympus-like 16-shot option. Nikon also offers a choice of Pentax-like Bayer-canceling or an Olympus-like res-boosting one, each with the option to perform it twice for a greater noise/tonal quality improvement. But, like Sony, these need to be combined off-camera, which requires a level of patience, file management and messing around in clunky own-brand software that builds into an appreciable hurdle.

Thanks to the complexity of the X-Trans sensor, the mode on its X-series camera requires 20 shots to deliver its pixel shift mode. Personally, I find that most subjects (even landscapes) have too much movement to wait this long and, on a grander scale, that life might be too short for such a commitment.

Limited benefits

I wanted to capture this burnt-out car, abandoned outside a defunct car dealership before it disappeared. I made the effort to arrive early on one of the only spring mornings with any light and then combined the images when I got home. As a result, I have a bit more detail, some odd cross-hatching in areas of movement, 16 massive Raw files and more information than I wanted about the lens’ corner performance. Yay. Sony a1 | Sony 16-35mm F2.8 GM @ 16mm | ISO 100 | F9 | 1/100 sec Photo: Richard Butler

Not only are they slow and clunky to use, but the benefits of multi-shot mode are often limited. Even with a static scene and optimal conditions, a 100MP multi-shot mode won’t match the results of a 100MP camera, but in most situations it often won’t even deliver its own maximal performance.

Unlike smartphones, which make extensive use of multi-shot combination, most large-sensor cameras read out their sensors quite slowly, creating appreciable delays between each shot, raising the risk of subject movement. The more sophisticated systems correct for this motion to some degree but do so by dropping back down to using a single image’s data, throwing away the detail benefit for any subject that’s moved, as well as leaving ghosts and artifacts around the image.

They’re also based on making precise sub-pixel movements, so are very sensitive to any camera motion, and can’t apply stabilization because the mechanism is too busy making pre-planned movements.

Finally, the flip-side of the higher resolution pushing aliasing to higher frequencies is that, just like a high-resolution sensor, the shots are quickly limited by diffraction. This means that you’ll need very sharp lenses, fairly wide open if you want to minimize the degree to which lens shortcomings and diffraction eat away at the hoped-for resolution boost.

My point being that it’s very easy to go to considerable extra effort for minimal gain. You still get the noise benefit, of course, but you can gain that by pressing the shutter button several times and merging the images yourself: you don’t really need a special mode for that.

Handheld multi-shot modes

Hand-held multi-shot modes don’t work the same way as tripod modes, so don’t offer the Bayer-canceling benefits or the same level of additional detail capture, but they’re usable in a much wider range of circumstances. This was shot on an impromptu hike on which I didn’t want to play tripod-sherpa. OM System OM-1 | 12-40mm F2.8 Pro II @ 18mm | ISO 200 | F5 | 1/1250 sec Photo: Richard Butler

Increasingly, we’re seeing handheld multi-shot modes appear, and these can be used in a broader range of circumstances. But it’s worth noting that these aren’t quite the same thing. Instead of moving the sensor in a precise, controlled way, they measure the degree to which your hand shake has moved the camera, then combine some of a burst of shots to try to boost detail levels. You don’t gain the Bayer-cancelling improvement in color resolution or sharpness through this approach, and won’t see the same degree of detail improvement.

Old man yells at cloud?

Ultimately, I’m not so vehemently against multi-shot modes that I don’t think they should exist. Even if they’re only useful to a tiny subset of users, I certainly don’t begrudge those people gaining a feature they want. But they’re so often so awkward to use and offer so little benefit in most circumstances, that I find it hard to be that enthusiastic about even the best (least-bad?) versions.

There’s a chance that my position is every bit as solipsistic as those who argue that cameras shouldn’t have video modes, just because they don’t use them. I’d like to think that my position is slightly different in that I dislike them because I don’t use them and have found them to be highly impractical and often awkward to use.

But to each their own. I’m certainly not about to start marking a camera down for having an extra feature, no matter how much clutter it adds to the menus. But equally, I’m not about to take up the cause of any commenters demanding that it’s a feature every new camera MUST have.