Paper: FlatCam: Thin, Bare-Sensor Cameras using Coded Aperture and Computation

Reference:

Asif, M. Salman, et al. “Flatcam: Thin, bare-sensor cameras using coded aperture and computation.” arXiv preprint arXiv:1509.00116 (2015).

Review:

A coded aperture based camera model is proposed by designing the aperture matrix in a row/column separable manner. The aperture matrix can be regarded as a outer product between two vectors with 1 and 0 only. The image formation process can be modeled as (1) $Y = AXB^T$, where X, Y, A, B are all $N \times N$ matrices. In the conventional image model: (2) $Y = CX$, where X and Y are both $N^2 \times 1$ vector, and C is $N^2 \times N^2$ matrix. Model (2) is much easier to solve by least square with L2_norm regularization compared with model (1).

The author put the aperture right in front of the CMOS chip. The space is 0.5 mm due to the glass. This can greatly reduce the thickness of the camera which might find its use in mobile phone camera.