3 ms·

The rise of computational photography is going to considerably change the camera landscape. This is already getting mainstream with the "depth effect" in the iP

by ot 10y ago

The rise of computational photography is going to considerably change the camera landscape. This is already getting mainstream with the "depth effect" in the iPhone 7+ and I believe that Google phones are doing something like that as well. Eventually the only thing we'll care about in a sensor/lens system (with possibly more than one of each) is the amount of information they can capture, and how fast. All the other parameters (sensor size, optical design) that currently influence the final photographic features will be simulated in postproduction, similarly to how today we can accurately simulate digitally the classical chemical films (Velvia, Tri-X, ...), enabling a great deal of artistic freedom. EDIT: I recommend watching this excellent presentation from the CTO of Light, a camera based on a sensor array, it goes into a lot of technical detail: https://www.youtube.com/watch?v=vKmC9xWHhM4 https://www.youtube.com/watch?v=vKmC9xWHhM4

ghaff 10y ago
I fully agree, although as with autonomous vehicles, it remains to be seen what actually happens in 5 years vs. what happens in 50 years.
- ot 10y ago
  Yeah but I think that the timeframe for smart cameras is a lot closer. Much of the theory behind computational photography has been well understood for a while and it's basically physical modeling and linear algebra. It just requires large amounts of computational power to work in realtime, and this is becoming feasible now.
mtw 10y ago
yes - but the same computational photography done on a Google Pixel can also be done on larger cameras. The Pentax K-1 for example can shift its sensor in 4 different directions for one single picture for increased dynamic range and color. So if one day you are able to simulate an amazing picture on a smartphone camera, with great deep learning filters, you could also create even more amazing pictures on a dedicated camera with a larger sensor. You'll just have to adapt the algorithms
- ot 10y ago
  That was my point: I'm not saying that phones will replace cameras for all uses (there are inevitable physical limits), just that cameras will look very different from what they look now.

by ot 10y ago

ghaff 10y ago

I fully agree, although as with autonomous vehicles, it remains to be seen what actually happens in 5 years vs. what happens in 50 years.

ot 10y ago
Yeah but I think that the timeframe for smart cameras is a lot closer. Much of the theory behind computational photography has been well understood for a while and it's basically physical modeling and linear algebra. It just requires large amounts of computational power to work in realtime, and this is becoming feasible now.

mtw 10y ago

yes - but the same computational photography done on a Google Pixel can also be done on larger cameras. The Pentax K-1 for example can shift its sensor in 4 different directions for one single picture for increased dynamic range and color. So if one day you are able to simulate an amazing picture on a smartphone camera, with great deep learning filters, you could also create even more amazing pictures on a dedicated camera with a larger sensor. You'll just have to adapt the algorithms

ot 10y ago
That was my point: I'm not saying that phones will replace cameras for all uses (there are inevitable physical limits), just that cameras will look very different from what they look now.