Upscaling (enlarging) digital images

William Jaeger

 

Simple beginnings: I wondered if a 36mp digital camera might come close to 4x5 in feel--the combination of tone and resolution and butter and grain that make large format what it is. All images are with a Sony A7r camera, July 2015.


First realization: 4x5 (Delta 100 film) scanned at 3200 dpi (or 4800 dpi) has far more dpi than a native digital file, even from 36mp cameras. But on looking, it seemed that the actual clarity, or feeling of detail and texture, was similar--or better!!--with the digital camera. I was probably scanning beyond the actual resolving power of the original film system (lens and film and developing combined). It gets truly complicated, but I’m zeroing in on how it looks. That’s the point, right?


CONCLUSION roughly--4x5 images have more grain and less apparent information than the unsharpened 36mp images, but the 4x5 images also have more acutance (a sense of sharpness). When the 36mp images are sharpened in various ways, they can seem to surpass the 4x5 images, but there is sometimes/always a tonal loss (a harshness, or lack of tonal range from area to area). I’m obviously not very scientific here. My gut has led me to want to continue with my 4x5 for now. It offers other advantages (like camera movements), and it is at least as good as the 36mp images I’m getting from my a7R.


Upscaling: so I decided to simply double the resolution of the Sony digital camera file. In Photoshop, I went to image size and doubled one dimension (with all the boxes checked below to keep proportions the same). Voila, an image that was the same “size” as my scanned 4x5. But how did it look? More on that in August.


For now: the BIG DIVERSION: my goal was to see if I’m doing this doubling, this upscaling, in the best possible way. Options are fairly few: 1) use Photoshop in one step as described with either bicubic or later options for preserving detail, 2) use Photoshop in a series of small  “iterations” where the scaling is done in reduced. repeated steps, 3) upscale straight from Lightroom with the Export mode, and 4) use an external program (which I don’t own, and which other sources say is similar to PS).


The bottom line: go ahead and use Photoshop in one step using image size to change your original file to the size you need. i used bicubic smooth. I tried careful iteration versions which have some support on the web and there was no visible difference. Not worse, not better. Nothing. (The logic of doing the enlargement in a series of small stages so that each step has a lot of real pixel information for each step is probably flawed in that those pixels are increasingly invented  and not “real” with each step.) The export straight from LR is identical to the one-step PS attempt as well, not worse not better. I’m guessing it uses the same exact processing steps. Why wouldn’t it? Anyway, I was a little disappointed but also relieved. Simple is good.


Upscaling itself is not the point here. It seems to smooth out jaggies and irregularities compared to just printing a small file big. So upscale in one step using the Image Size window in Photoshop and everything will be optimal, as far as I can tell.


EVIDENCE for what it’s worth is on this next page (click!)