More than one person has asked me about what the heck I mean by pixel shift and why we would want it (I list it in my Z System Wish List).
Pixel shift means using the VR capability on the sensor to move the sensor location a small amount between shots. Generally, two basic forms exist, each with a different benefit:
- Four images, each shifted one pixel — A shot is taken, the sensor is shifted right one pixel and another shot is taken, the sensor is shifted down one pixel for the next shot, then left for the last shot. In this form of pixel shift you get two primary benefits: noise reduction and removal of color aliasing. That's because at each photosite location we get RGBG values, not just a single R or G or B value. Some talk about additional resolution, but it's really the removal of aliasing they're seeing, which produces better edge definition.
- Eight images, each shifted a half pixel — After the first shot, you take an additional seven images with the sensor shifted a half pixel each time (across across, down down, back back, up). While there's again some noise reduction benefit, the reason to perform pixel shift this way is to obtain additional spatial resolution. While it's not a perfect correlation due to the overlaps, many think of this as a "double the resolution" trick.
Other variations exist. The Sony A7R Mark IV can perform a 16-image shift. That produces a 240mp result.
The drawback is that the subject(s) can't be moving for the shift to work, and you want the camera on a sturdy support during the sequence. Thus, the primary beneficiaries of such an addition would be landscape, architectural, macro, and some travel photographers. Another drawback is that you normally have to process the raw files in some way to get the aggregated effective image. In Sony's case, their software is a bit of a pain to use; most of us use my friend Iliah Borg's PixelShift2DNG when working with Sony pixel-shifted raw files.
So why is this feature on my wish list? Because competitors have it, and it does open up new avenues of use for our cameras, particularly for landscape and architectural shooters.
If Nikon is reading this — here's the thing that none of the camera makers are understanding: we often want to combine these "more than one shot" ideas together. For instance, imagine pixel shift with focus bracketing. Or even just with exposure bracketing. While the engineering teams see the opportunity for a new feature via its technical nature, they're not quite getting how the user will interact with it or want to set these functions up. So while competitors may offer the pixel shift feature, it tends to be a geeky, must know what you're doing type of feature, not something you just dial up and press a button to use. There's plenty of room to make pixel shift a more useful feature than others have given us, which is another reason why Nikon should pursue it.
