| Olympus and Sony have been the most aggressive about sticking clever software in their cameras, so they'd be the ones most likely to invest more in computational photography. The problems for camera manufacturers around this, though are: 1. Their market is incredibly, loudly regressive about a lot of this stuff. A noisy chunk of the photography market is really hostile to workflows that don't mimic hundred year old darkroom processes. Doing in-camera is an abomination. Automation is an abomination, etc. 2. Building the compute in to the camera is non-trivial. You've already got a lot of compute power focusing on running the complexities of things like continuous AF tracking (e.g. on top-end Olympus that's 4 of the 8 cores available). A lot of compute budget is used for things phone cameras are rubbish at. At some point things like heat become a problem. 3. Data volumes are hard: Olympus are "only" dealing with a 20 MPix image. Sony are dealing with 24 - 60 MPix images. Olympus do that at up to 60 frames/s (20 with AF), and to read the data within the constraints of shutter speeds of 1/16000 s. That is... a lot of data compared to the relatively modest rates on an iPhone. Oh, and while iPhone users are generally OK with a bit of lag, DSLR users get really pissy if you're making them wait for that. Latency needs to be very low. 4. Physics is hard: A Sony sensor is a 24 x 35 mm hunk of silicon. Quite apart from the challenges of the data volume, reading a chunk of CMOS that large has been a challenge. Sony have done a lot of clever things to work around those limits, but still... (Olympus have been able to do high frame rates longer than anyone else in part because their sensors are smaller) |