3D Cameras for HDR

摘要

High dynamic range 3D images are generated with relatively narrow dynamic range image sensors. Input frames of different views may be set to different exposure settings. Pixels in the input frames may be normalized to a common range of luminance levels. Disparity between normalized pixels in the input frames may be computed and interpolated. The pixels in the different input frames may be shifted to, or stay in, a common reference frame. The pre-normalized luminance levels of the pixels may be used to create high dynamic range pixels that make up one, two or more output frames of different views. Further, a modulated synopter with electronic mirrors is combined with a stereoscopic camera to capture monoscopic HDR, alternating monoscopic HDR and stereoscopic LDR images, or stereoscopic HDR images.

主权项

1. A method comprising:
receiving a left input frame created with a first luminance range and a right input frame created with a second luminance range, the left input frame comprising a plurality of left input scanlines, the right input frame comprising a plurality of right input scanlines, each individual left input scanline in the plurality of left input scanlines corresponding to an individual right input scanline in the plurality of right input scanlines, and the first luminance range being different from the second luminance range; determining a first left input scanline in the plurality of left input scanlines and a first right input scanline in the plurality of right input scanlines, the first left input scanline corresponding to the first right input scanline, the first left input scanline comprising a first set of left input pixels, and the first right input scanline comprising a first set of right input pixels; selecting a first subset of left input pixels among the first set of left input pixels; selecting a first subset of right input pixels among the first set of right input pixels; computing, with the first subset of left input pixels and the first subset of right input pixels, first input scanline disparity between the first subset of left input pixels and the first subset of right input pixels; generating, based at least in part on the first input scanline disparity, a first left output scanline for a left high dynamic range output frame and a first right output scanline for a right HDR output frame.