Reader Comments: Frame Averaging
Claude F writes:
Been following the reviews. I’m interested in the exposure blend process, particularly using only two exposures. How exactly is this done and what is the post process process. You may have this info on the site but I could not find a clear and concise description. What I did see was unclear, to me at least.
DIGLLOYD: agreed that I have not made it clear how to do frame averaging. It is on my to-do list to add a section in Making Sharp Images on how best to do frame averaging, both shooting and processing.
In brief, stack all the layers in Photoshop, then set the opacity of layer L of N to {100%, 50%, 33%, ..., (1/N)%}. For example: {1, 1/2, 1/3, 1/4, ..., 1/N} or in other words 100%, 50%, 33%, 25%, 12%, etc.
It is also my intent to make available a Photoshop script to automate the task including aligning the frames (if needed), setting the layer opacity, then generating one averaged layer from all of them. I already use such a script, but I need to tweak it and make it standalone for my readers). With the script I use, it is trivial to do it in seconds with one click.
Richard R writes:
My Sony A7Rlll does not support pixel shifting.
I am wondering if the quality you achieved by "frame averaging" can be achieved through PS and "stacking" images for either focus or exposure differences? Just take multiple images... I haven't found any discussion of "frame averaging" except in technical journals.
DIGLLOYD: see above for how I do frame averaging in Photoshop. Some cameras like the Nikon D850 and Nikon Z7 can average frames in-camera, producing an averaged raw file. But (tedious and frustrating), only two files at a time, so to average 4 frames, average each pair, then average the two average frames.
The Sony A7R III does have pixel shift. Program it into the for easy access.
However, I am skeptical that pixel shift has the required accuracy—I see over and over in the field that pixel shift has pixel skew under constant lighting conditions given the image-wide checkerboarding. Basically, my conclusion is that pixel shift has some slop factor in it, and that’s why it works so poorly in general for all purposes. My guess is that pixels do not align exactly. Or it may be that different rows of the image are just not adjusted properly for overlay purposes (very common in Bayer matrix sensors). Thus pixel shift is convenient, but might not be an ideal choice for frame averaging.
The best way to do frame averaging is electronic shutter plus remote release. By so doing, there is absolutely no movement with the camera. With the Sony remote, I can shoot 4 frames in under a second with 4 presses (avoid burst mode as it records 12 bit files).
The subject can still move however, which make frame averaging problematic for many subjects (any wind pretty much rules it out). However, one can do it and mask off problematic areas, or mask in frame-averaged areas.
Indeed, one can combine frame averaging with multiple exposures, masking the differing parts as shown below. This image is two exposures overlaid.