SiliconValley.com published a report from the meeting with Pelican Imaging CEO Chris Pickett and VP Marketing Paul Gallagher. Few interesting statements:
- Pelican's software takes the images recorded by each of the 16 sensor arrays and combines them, yielding not only an 8MP image, but one with depth information for each point within it.
- Pelican Imaging technology could allow users to easily select an object in an image to adjust its exposure or copy it to another picture. It could allow users to interact with their phones with 3-D gestures. And it could be used in for face-detection systems that determine whether someone is authorized to use a device.
- Pelican says that it's already working with some of the existing smartphone camera system makers and hopes to have its system on a smartphone by early next year.
 |
| Pelican Imaging Camera |
Indeed, this is much more than Ultrapixel from HTC. Imagine that you cover the back of Iphone with, say 1024 cameras. Then if the noise is random, then you can improve 32 times the S/N ratio. You sell this concept to DoD and it's more credibale than SiOnyx. A high quality CMOS sensor can see under moonlight, so 1024 cameras can see 32 times less than moonlight = starlight.
ReplyDeleteThe only thing that ultimately matters in an imaging system when figuring SNR under a particular illumination level is the number of photons captured. This, in turn, depends only on the area of the first element in the optical system. What you do with the captured photons ia a decision that is embodied in how the optics spread the captured photons over and image plane, the size of the image plane and how the image plane is divided into pixels. With that information, the photon flux per pixel can be calculated. Then, the degradation in the photon image can be determined based on the QE and noise of the image sensor.
ReplyDeleteWithout stating all of these, camera comparisons are meaningless.
what do you want to say please???
DeleteSimply that the active area of circles inscribed in a square is inferior to the active area of one circle circumscribed in a square (look at the picture of the phone). Basic logic.
ReplyDeleteNoise is proportionnal to captured flux...
They're equal actually. Imagine shrinking your circle down to a 1/16 of the size and then having 16 of them, that will give you the same area. The advantage of this method is that a larger senor (one physical or 16 smaller ones) can be used without increasing the length of the lens which is important for smart phones. Then there's the depth information which has numerous uses. Unless there's a cost or complexity catch I can see this being quite popular on phone.
DeleteWho is going to pay for such an expensive front facing cam while today they go for 1-2 usd ?
ReplyDeleteThere is a catch: having 16 small size image does not sum up to one high res image so you loose a factor of 16 on the resolution. You do gain depth but at a high price. maybe with some super res you can improve by 50% or so but still you loose a factor of 7.
ReplyDeleteAlthough the press release says an 8mp image. I imagine there will be artifacts but to be honest even 2mp is probably fine for 99% of cellphone images.
Delete