I recently sold a camera (Nikon Z50) and bought a new camera (Nikon Z6III), mainly because it has much better auto-focus and high-speed shooting performance for wildlife. So far I am very happy with the auto-focus and speed performance as it is everything I expected. But it also has another feature that I have been curious about, and that is pixel shift. Many camera brands have this feature, but I am looking at the implementation on a Nikon Z6III in this blog post.

Pixel shift is a feature that allows you to set the camera to automatically take multiple photos, shifting the sensor slightly for each image. These images can then be merged using the Nikon NX Studio software to create a higher resolution image. In the case of the Nikon Z6III the standard image is 6048 x 4032 (about 24 megapixels) and an image created from merging 32 images is 12096 x 8064 (about 97 mega-pixels). So, theoretically you would expect four times the detail in a merged image. But I had my doubts, mainly because I don’t think the lenses are that good. So I have performed some experiments not because I want super high resolution images, but because I am curious.

First, a scene at the lake taken with the Nikon Z6III with a 24-200mm f/4-6.3 lens. This is one of the value lenses and not the highest optical quality. The image below is the merged image, but not exported at full resolution because it is a 517MB file.

Next I have crops of the single frame image and the merged image for comparison. On the left is the crop of single frame image (636 x 636) and on the right is the crop of the merged (1272 x 1272) higher resolution image. Looking at the texture of the tree bark, I think the single frame image looks better even though it has a fourth of the resolution of the merged image. I am a bit confounded by this as I was expecting the opposite to be true.

Maybe it’s the lens. Trying to get that sort of resolution out of the 24-200mm superzoom is pointless. When I was at the lake that morning, I didn’t have my 35mm f/1.8 prime with me and that is the sharpest lens I own. So, a few days later I made another experiment using the 35mm lens this time. I made certain that my tripod was stable, there was no wind to shake the camera, I put in a 2 second delay so that I wouldn’t cause camera shake. This should be ideal conditions for a pixel-shift image. Below is an image of some boulders taken with pixel-shift; 35mm, f/8.0, 1/10s, ISO100. Yeah, the sky is blown out, but I was more worried about the boulders and not about shooting HDR.

Below are some crops of the boulder that I set focus on, side-by-side for comparison. On the left is a single frame and on the right is the merged image. All of the edits and crops were copied onto each image. Still, the single-frame is better in my opinion. Look at the lichen on the boulder; much more detailed in the single-frame image.

I didn’t understand these results. I wasn’t expecting to be wowed with the improved resolution, but I was expecting the results to be marginally better. I haven’t found much online about people doing similar experiments. Assuming that the results should be better, I was either doing it incorrectly or something is wrong with my camera. I tend to believe it is my technique as Nikon makes a darned fine camera.

One option that I didn’t try was adding a delay between photos as I didn’t consider it necessary, but there is a chance that this poorer image quality is caused by shutter-shake. There is the option to add a delay between photos, so I tried this using my bookcase as a subject in my wind-free office. First I ran the pixel mapping function, like the manual suggests, though this is ostensibly to reduce the impact of ‘hot pixels’ but I figured it couldn’t hurt. I used the same 35mm f/1.8 lens and set the aperture to f/4, which is where it reaches peak center-sharpness according to what I have read. I put the camera on a tripod, set it to take 32 pixel shifted images, with a 2 second start delay and a 1 second delay between images. I then pressed the shutter button and sat back and let the camera do its thing. So with all that, I have the single-frame (left) and merged image (right) below for comparison. I focused on the center of the frame, which were the letters “Ernest Hemingway” on the book binding. What do you think?

So, at last I think I have a higher resolution image that is at least as sharp as the single frame image but with greater resolution. If you keep zooming in to the pixel level, you do get four times more detail in the merged image. I think the key point was adding a delay between images to minimize shutter-shake. I suppose you could also use a really fast shutter-speed, but you would probably have to up the ISO and get more noise which would defeat the purpose of a high-resolution image.

So, in conclusion, I think it is possible to get good results from this feature if you have plenty of patience, a really good lens, and an extremely well-behaved subject; but for most situations it is probably not worth the effort. I didn’t care anything about the pixel-shift feature when I bought this camera, so this is just a curiosity to me. The features that I do care about, the animal autofocus and high-speed shooting are amazing with this camera and it writes to the memory card so fast that you cannot fill up the buffer. It is a bit heavier than the Z50, but it is still a good hiking companion for when I want to travel lightly.

If your camera doesn’t have this feature but you must have gigantic file sizes, there is also the option of upscaling an image in Photoshop. Though now you are adding in data that wasn’t actually present in the scene but rather estimated by an algorithm. I have done this before and found the result to be unimpressive. Besides, I have a large 36″ x 24″ print hanging on my wall right in front of me that was shot with a 24MP camera and it looks great even if I stand 6 inches from it (which is a ridiculous thing to do).

Thanks for reading. Let me know what you think below?