| 摘要 |
Determining a first head orientation, selecting two image sources to form a first stereo image pair, and rendering a first stereo image by rendering one image of the first stereo pair for each eye, then determining a second head orientation, selecting two image sources to form a second stereo pair wherein one of the image sources is the same source used as part of the first stereo pair, and rendering a second stereo image by rendering one image of the second stereo pair for each eye. Virtual eye locations may be used alongside closeness criterion. Image stitching or blending may combine images with a temporal transition based on head speed. Stereo audio may be used. Image sources may be real-world cameras or virtual cameras for producing rendered images. Encoding stereo video using a plurality of cameras with active and passive scene areas and different fidelity is also defined, as is a camera for creating 3D video comprising at least three cameras, preferably eight wide-angle cameras, having overlapping FOV. In the present invention, when the user tilts their head when using a head-mounted display (HMD), a new stereo pair is formed by switching to the best camera for realistic stereo disparity. |
