A 3MPixel Multi-Aperture Image Sensor with 0.7μm Pixels in 0.11μm CMOS

Conventional image sensors have improved with technology scaling mainly by reducing pixel size to increase spatial resolution [1,2]. As resolution approaches the limits of existing optics, is there much to gain from further pixel scaling? In [3], we argue that further scaling can provide new imaging capabilities via a multi-aperture (MA) architecture, which consists of an array of small submicron pixel imagers (apertures), each with its own integrated optics. By focusing the integrated optics onto an image plane formed by an objective lens in a region above the MA imager, the apertures capture overlapping views of the scene. The correlation and redundancy between apertures, along with computation, provide several new capabilities, including: (i) simultaneous capture of a 2D image at higher resolution than the aperture count and a 3D depth map without the need for active illumination or calibration; (ii) simplification of the objective lens design; (iii) reduction of color crosstalk via per-aperture color filters; and (iv) increased tolerance to pixel defects. It is further shown that depth resolution continues to improve with pixel scaling beyond the typical spot-size of the optics. Designing scalable arrays of submicron pixels with acceptable imaging performance is challenging however. In [3], we propose using a frame-transfer (FT) CCD aperture with CMOS readout architecture. In [4], we demonstrate a 16×16, 0.5μm pixel-pitch FT-CCD in 0.11μm CMOS technology with acceptable imaging performance.