Waskul.tv publishes a 18 min-long interview video interview Pelican Imaging’s Founder and CTO Kartik Venkataraman where he discusses strengths and weaknesses of the company's array camera technology.
again. I guess the investors are angry. complete failure in mobile. quality of pics are lousy compared to the standards of today. I think their days are numbered. VP biz dev left and joined OVT. That usually means the end.
The 16 camera prototype seems to be a very low end array using only one small sensor totaling 12 Mp, but i think it should be easily applicable to better hardware.
I just cant wait for getting my hands on a phone with an array camera with many high end 1/3" sensors combined with good software.
Isn't it cheaper to produce many small sensors/lenses then one big sensor and lens with the same sensor area and aperture? I'm not shure where the cost of largeness starts skyrocketing.
Aside from that, thin devices is a selling point of its own. Nokia Pureview 808 1/1.2" sensor with f/2.4 was pretty thick for a phone. If someone made a phone with four 1/2.4" f/2.4 cameras it would be the same theoretical sensor area and aperture area. If that would cost the same it could sell better because it would be thinner.
Several camera modules can also be used for a variety of interessting features like 3D, depth awareness, alternating video frames (multiplying the framerate by the number of cameras in the array), HDR features, multiple real focal lengths and so on.
Maby even do something controversial as making a mirrorless camera competitor with APS-C equivalent sensor area or bigger (large array of smaller sensors and lenses), in a phone like form factor.
It's not clear to me that it's cheaper to make many small sensors in an array (very large silicon area) than a medium sized sensor. For optics, it certainly is cheaper to make small optics.
However, from Pelican Imaging's tech docs, it looks like there may be limitations on the resolution of their optics. I would be surprised if they're able to make individual small lenses at 5-8mpix resolution. If they can't, then how are they able to get 5-8mpix output? Super resolution can only turn some data in the form of aliasing into meaningful data, but if the data isn't there to begin with...
Also, the accuracy of the super resolution depends heavily on the accuracy of the disparity map. For 8mpix resolution from many .75, they would need >1/3? pixel accuracy from their correspondence algorithm. It definitely doesn't look like that's what they're getting based on the published depth maps.
" If someone made a phone with four 1/2.4" f/2.4 cameras it would be the same theoretical sensor area and aperture area."
Fusing 4x 10mpix from an array doesn't give you 40mpix. The specs wouldn't be the same.
There is no doubt that there is value in having multiple apertures. This is why there are so many players in this space -- even stereo gives you something cool.
It's not clear to me that the approach Pelican Imaging is taking is optimal.
Pelican will likely sell itself to a major OEM (for peanuts) to recover investor money. Otherwise, there is no future for Pelican Imaging at any major phone/tablet OEM and they will simply cease operation.
I'm all for pushing new technologies, but Pelican was nowhere ready with their technology for the big show. There are so many shortcommings of the tech which are still unsolved.. just the artifacts on their latest pics are astounding... I don't understand this.. why try to sell an airplane that can't fly?
again. I guess the investors are angry. complete failure in mobile. quality of pics are lousy compared to the standards of today. I think their days are numbered. VP biz dev left and joined OVT. That usually means the end.
ReplyDeleteThe 16 camera prototype seems to be a very low end array using only one small sensor totaling 12 Mp, but i think it should be easily applicable to better hardware.
DeleteI just cant wait for getting my hands on a phone with an array camera with many high end 1/3" sensors combined with good software.
Sure, if cost were no object, the technology works. But cost is important for commercial success.
DeleteIsn't it cheaper to produce many small sensors/lenses then one big sensor and lens with the same sensor area and aperture? I'm not shure where the cost of largeness starts skyrocketing.
DeleteAside from that, thin devices is a selling point of its own. Nokia Pureview 808 1/1.2" sensor with f/2.4 was pretty thick for a phone. If someone made a phone with four 1/2.4" f/2.4 cameras it would be the same theoretical sensor area and aperture area. If that would cost the same it could sell better because it would be thinner.
Several camera modules can also be used for a variety of interessting features like 3D, depth awareness, alternating video frames (multiplying the framerate by the number of cameras in the array), HDR features, multiple real focal lengths and so on.
Maby even do something controversial as making a mirrorless camera competitor with APS-C equivalent sensor area or bigger (large array of smaller sensors and lenses), in a phone like form factor.
It's not clear to me that it's cheaper to make many small sensors in an array (very large silicon area) than a medium sized sensor. For optics, it certainly is cheaper to make small optics.
DeleteHowever, from Pelican Imaging's tech docs, it looks like there may be limitations on the resolution of their optics. I would be surprised if they're able to make individual small lenses at 5-8mpix resolution. If they can't, then how are they able to get 5-8mpix output? Super resolution can only turn some data in the form of aliasing into meaningful data, but if the data isn't there to begin with...
Also, the accuracy of the super resolution depends heavily on the accuracy of the disparity map. For 8mpix resolution from many .75, they would need >1/3? pixel accuracy from their correspondence algorithm. It definitely doesn't look like that's what they're getting based on the published depth maps.
" If someone made a phone with four 1/2.4" f/2.4 cameras it would be the same theoretical sensor area and aperture area."
Fusing 4x 10mpix from an array doesn't give you 40mpix. The specs wouldn't be the same.
There is no doubt that there is value in having multiple apertures. This is why there are so many players in this space -- even stereo gives you something cool.
It's not clear to me that the approach Pelican Imaging is taking is optimal.
Pelican will likely sell itself to a major OEM (for peanuts) to recover investor money. Otherwise, there is no future for Pelican Imaging at any major phone/tablet OEM and they will simply cease operation.
ReplyDeleteI'm all for pushing new technologies, but Pelican was nowhere ready with their technology for the big show. There are so many shortcommings of the tech which are still unsolved.. just the artifacts on their latest pics are astounding... I don't understand this.. why try to sell an airplane that can't fly?
ReplyDelete