Rice University reports no-lens camera
The Flatcam has similarities to lensless image sensor being developed by Rambus, which replaces the lens in a conventional camera with a diffraction grating and computation (see Rambus develops lensless image sensor for IoT). However, the Flatcam does provide a complete optical reconstruction of the given scene, while the Rambus system does not as it is intended for use by machines.
The advantage of the Flatcam over a conventional camera is its extremely thin profile. As the coded mask is positioned almost directly on top of the sensor the assembly can measure as little as 0.5mm. The disadvantage is the energy cost of computation required to recreate the image.
Every point in a scene casts an image on the sensor, but this data must then be processed using the reverse function to that represented by the mask to produce the image. Source: Rice University.
The mask consists of a pattern of opaque and transparent areas and each transparent region can be considered as a pin-hole. Light gets diffracted and modulated by the mask such that light from each point in the scene casts a complex shadow on the sensor and this mapping can be represented by a linear operator. A computational algorithm can be deduced and used to recover the original scene from the sensor measurements.
The potential of the architecture has been demonstrated using a prototype camera made using a commercially available sensor at 512 by 512 pixel resolution and a mask. The monolithic integration and extreme thin-ness could make such cameras useful for mobile and wearable applications and the Internet of Things, the authors claim.
The paper has been accepted for presentation at the Extreme Imaging Workshop that is part of the International Conference on Computer Vision (ICCV 2015) due to take place December 17, in Santiago, Chile.
Related links and articles: