Wednesday, May 20, 2015

The Future of Imaging

I got into it a little with Michael Reichmann the other day, arguing about what Camera Companies Ought To Do (like all pundits, Michael thinks they should build a special camera just for him). It got me to thinking about what the future will be like.

Not next year, but a few years from now. How would you get out in front of the market?

Here's my theory. Free business idea, if you want to take it on.

The problem is one of curation. We have no problem with tons of photos. Making lots and lots of sharp detailed pictures is a solved problem. This has created new problems, of curation and of preservation.

We are swamped in pictures. We are swamped in our own pictures, often. The current solution is to simply let the older ones go. They slide down some timeline, and get looked at less and less. Even I do this, and I am incredibly old school, and curate like the devil himself. I easily can get to photos from 5 years ago, 8 years ago. I don't, though, not very often.

So let's pose some problems that a typical person, someone who as I like to say just wants pictures might run in to.

  • I miss grandma.
  • Wow, the kids grow up so fast.
  • Do I have any pictures of my car before the accident?
  • What color was dad's house?

These can all be solved by something which can search for relevant files, sort them, organize them, and present them quickly and easily. Something that lets us easily:

  • select
  • organize
  • render
the right pictures, at the right time, from multiple large archives of pictures.

Selection is obvious. It should use, at least: face, scene, and object recognition, accompanying text, embedded date/time information, context (if any -- at least the other pictures nearby both in the sense of location of the file, and in the sense of time).

Organization should be similarly obvious, it's simply how we arrange the pictures (not necessarily in a linear sequence, by the by), and we need good tools for quickly and easily doing that. But this is a more or less solved problem, we know how to build tools for rapidly flicking through pictures, dragging, dropping. Still, automated solutions that give an excellent first draft are pretty much a requirement here.

Rendering is less obvious. We might want a movie with Ken Burns effects, we might want to populate a digital picture frame, or our computer's desktop background. We might want to print a book. We might want to show a friend, or share a slideshow or a movie. We might want to create a composite portrait. Queries like "what color was" might be simply a set of color swatches representing best estimates, with clickable context. We might want a 3D printed bust of grandma? I dunno. The possibilities, if not endless, are large.

It's an indexing problem, an image recognition problem, a big data problem. Ideally we'd have some automated simple composition tools -- some basic crops, color correction. Are we looking for portraits of grandma? Then make portraits of grandma out of whatever raw material is present.

Many of us have piles of photographs lumped into flickr, facebook, picasa, local hard drives, and so on. Many of us will have bigger piles as time goes on. It's obvious, I think, that any tools need to be able to go dredge out photos from all these sources and more, and should support plugins for arbitrary archives.

If we could solve the problem of managing immense heaps of pictures, then we could go really crazy. We could all wear wide-angle lapel cameras taking a picture every second, dumping it all into the cloud, and the result would no longer be a incomprehensible mess.

Now we can say "make me a shutterfly book of Susanne's birthday party from last week" and then we quickly flip through the mockup, make a few light changes, and press PRINT. Done. The software should find all the photos from the right timeframe that look "birthday party". It should identify all the unique people. It should construct a set of pleasing portraits with suitable croppings of suitably selected photos, and it should pick out a decent sequence of:

  • the piƱata
  • the present opening
  • the cake cutting and candle blowing

based on some expert system's understanding of what local customs for birthdays are, and what they tend to look like.

Perhaps wrap up with a couple of collages.

I don't think the future of "imaging" is in imaging at all, although it may lead toward specific devices eventually. Lytro cameras, anyone? It would be trivial to integrate the ability to retask a given file as a shallow DoF portrait, or as a record shot of Sue At The Party. This, again, could be automated. Are we looking for portraits of Sue, or are we making book of The Party?

I think the future is in managing how we "use" pictures.

No comments:

Post a Comment