In the smartphone versus dedicated camera comparisons I keep seeing, there's an interesting sub-theme that everyone seems to be missing. It actually corresponds to why I decided to write Mastering Nikon JPEGs (due in Q1 2025).
Put simply, everyone is comparing smartphone "build-a-scene" with camera defaults.
What do I mean by this? Let's start with "build-a-scene." Almost all smartphones today run with the sensor actively before and after any "shutter release" action. They take anywhere from eight to 32 images and run the image processor on all of them to create the "moment" you pressed the shutter release, but with additional information from the sub-frames. This is how smartphones look less noisy than their sensors are: they're stacking static portions of the image to remove the photon randomness caused by how little light got to the image sensor.
The latest and greatest smartphones do even more machine-learned processing than that; it's one of the reasons why Apple started putting Neural processing cores in Apple Silicon: to speed that up. Edge detection, motion detection, and a host of other detections are all done on the full set of pixels before producing a final image. The so-called Portrait modes first subject detect, then process the subject and background frames differently.
Some people call this computational photography, though most of the algorithms I've seen are really just machine-learned post processing done in real time. In the end, both iPhone and Android devices are essentially doing one heck of a lot of pixel pinching, pushing, crunching, inventing, and more.
Could you do the same with say, a Pixel shift image stack in a dedicated camera? Absolutely.
Meanwhile, the dedicated cameras in pretty much all of the image comparisons I've seen are essentially "taken at camera defaults." The reasoning behind this goes way, way back (20 years or more). Essentially, it boils down to "I guess this is what the camera makers wanted us to see." Virtually every reviewer or comparison maker probably thinks they could create better out-of-camera images than the camera defaults, but they don't have the balls to try that in public.
What the camera makers have always had to do is make devices that move to the mean for an average user that doesn't want to spend any time thinking about setting things. Mostly because it starts to get too complicated to do anything else (remember, even the smartphones are using AI/machine learning to do it at all).
So what's really being compared is Auto Processing (smartphone) versus Auto Settings (cameras).
Frankly, I think the camera makers missed a beat or two along the way. That's particularly true of fixed lens cameras, but it's become true of mirrorless cameras, as well. If the image sensor is running all the time, as it is in those two cases, you'd think that you could build a better understanding of the scene that you can apply when the photographer actually presses the shutter release. Instead, it appears that almost all dedicated cameras look briefly at the most current information, make simple adjustments, then punt that data if anything changes (focus, setting, composition, you name it). The most recent single sample taken prior to the shutter release seems to be what generates white balance, exposure, and color decisions.
One thing I noticed once Apple started letting us iPhone users get to the original data (or at least a subset of it) was that I could get better-looking images out of my smartphone than all of Apple's intelligence could. Hmm. That corresponds to what happened when the camera makers gave us raw image formats ;~).
Unfortunately, most of the customers using a camera of any sort, perhaps nearly all of them (>90%), want automagic, a word I invented something like 40 years ago. In other words, press a button and the machine does all the thinking, setting, and rearranging for you. What is really being compared in every one of the smartphone versus dedicated camera comparisons I've see to date is "how's the automagic work?" I'll give the gold crown here to Apple first and foremost, with Google a step behind, and Samsung, et.al. right there at their tails. Dedicated cameras bring up the rear.
Funny thing is, Nikon at one time worked on this problem with Coolpix. With things like Best Shot Selection they were doing exactly what the smartphone makers jumped on later: don't just look at one frame! Nikon also deployed a version of Live Photo (something Apple also later added) as well as a bunch of other things. I can't say anything more here due to NDAs, but at one point I was hired as an expert to do patent search and analysis in exactly this area. It's interesting that the camera makers didn't defend a turf they had already started exploring.
Unfortunately, the camera makers are now in a tough situation. The R&D cost of full battle with the smartphone capabilities while running an image sensor constantly and improving that with neural engines doesn't spread well among the few remaining units the Japanese are selling of dedicated cameras these days. In actuality, even the phone makers are grappling with that same problem now that smartphone sales have stalled on a volume plateau far above the dedicated cameras. One reason why you're seeing Apple pour so much of each iPhone generation difference into the camera side is that it's still the one point where they can clearly differentiate and improve compared to other smartphones, but you have to wonder how long those legs are. Apple's marketing has already turned to AI as the new differentiator, for example.
So what's Thom's Maxim here?
Thom's Maxim #237: If you have all the data, you can always do better than a built-in automated process.
Yes, automatic features are nice to have. No, they don't produce the best possible results.
