As someone who does research on inverse signal recovery, what can be recovered nowadays is absolutely mindblowing. I work on the group synchronization problem, which aims to extract an underlying ground truth signal from noisy, shifted copies of that signal. In my case, I am extracting atomic environments from noisy rotated and permuted copies of an "archetypal" atomic environment, but I imagine the same technique applied to photography would produce stunning results. You would be surprised with the amount of noise for which you can still recover the true signal.
See https://arxiv.org/abs/1607.03464 for an idea of what's become possible recently.
The problem I see with that though is
1) it's still too big 2) it's still too expensive and 3) I bet it adds an enormous amount of complexity to manufacturing
Lytro had really awesome technology something like 10 years ago for a camera that could focus on anything in the image (it didn't take pictures exactly, which was one of the problems, how do you manage the files which also happen to be gigantic). Awesome idea, but it just didn't work in the consumer market.
I would LOVE to see Light succeed and have a substantially better camera integrated into smartphones, and would even pay a few hundred dollars premium for it, but if that happens at all it's going to take years and I doubt many people will be happy to pay that much more when they just want to use Instagram.
- the demo photos, like you mentioned have either a softness like you mentioned or colors that seem a bit too vibrant. Would like to see more examples of action photos, up close macros, and night shots.
- no stabilization - in their talk they mentioned this isn't practical and they'll sell a stick gimbal, though it seems odd when a smartphone has this.
- the delays, it took about a year longer than I initially expected (preorders are shipping this July)
- My own needs - I'm realizing I'm not that nuts of a photography enthusiast lately for the price of the camera.
There's also some hesitance when reading the founders mit technology review article where he hints that the end user will do more things in post with their first product until they leverage an fpga/asic for processing (I think they're currently using a snapdragon 820 with an fpga for triggering)
Still rooting for them though since I hated carrying even a micro 4:3 camera
They're working on a light field camera and stack that they advertise will allow the viewer to don a VR headset and view a live-action scene with 6 DOF, so long as the viewer doesn't move their head terribly far.
I hope the technology is awesome, but based on what I've seen (and heard from employees) it's not going to live up to the hype.
As long as bigger cameras perform better (and there's a demand for better performance) they will keep growing until they take up the entire back of the phone.
My bet is on micro lenses so small that the camera becomes a flat rectangle taking up most of one side of the phone, along with some crazy insane software processing.