Prop research: combining frames into superresolution image

[Maniacy]

If you want to have a very close look at some props in a movie, you can use this trick to combine a number of consecutive frames into a single image of higher resolution and detail.

I've used this tutorial:
http://petapixel.com/2015/02/21/a-practical-guide-to-creating-superresolution-photos-with-photoshop/

with 80 frames taken from the movie "Fear and Loathing in Las Vegas" because I want as good an image as possible of the case and it's contents. All images contain Duke's hand. Here, the first and last image I used.



First, I loaded all screenshots as multiple layers like described in the tutorial above. However, then I applied a mask to each single layer and manually masked out Duke's hand.
Then I continued with the process:

1. (Import all photos as stack of layers) - done
2. Resize image to 4x resolution (200% width/height)
3. Auto-align layers
4. Average layers

As averaging method I converted the stack into a smart object and then set the blending method to 'median'.
Afterwards I used the Smart Sharpen filter as described and used 'Curves' to brighten the image a little.
This is the result: (download and open to see full resolution)


It doesn't work wonders, but it's a marked improvement over looking at the individual frames.

 

[Mike J.]

You can get a similar result by playing with the Unsharp Mask filter. I will grant that your composite is 'cleaner' in the more digitally compressed (dark) areas.

 

[Maniacy]

Thanks for pointing that out.
The results may look the similar but this method really combines the information present in the 80 frames. The estimation of where a given contour is is more accurate than the unsharp mask filter and you get an image that will look good in print. And of course, you can use the unsharp mask filter on this image and get an even sharper result.
Granted, this is not the best example because the 80 frames are all very similarly unsharp. It works even better when there's motion blur. Try it out. This method is based on statistics and has scientific valor.
Also, it perfectly hides things moving through the image background.