Processing image data from multiple cameras for motion pictures

摘要

An approach is disclosed for generating stereoscopic image sequences. The approach includes receiving a plurality of image sequences, each image sequence captured from a distinct camera or other capture device in a multi-camera array. One image sequence is a primary image sequence and the other image sequences are secondary image sequences. The approach further includes generating, from the primary and at least one or more of the secondary image sequences, depth maps corresponding to images of the primary image sequence and generating, for each image in the primary image sequence, based on at least the depth maps and the corresponding images of the primary image sequence, stereoscopic images of a stereoscopic image sequence.

主权项

1. A computer-implemented method for processing image data, comprising:
receiving a plurality of image sequences, each image sequence captured from a distinct physical camera in a physical multi-camera array, wherein the image sequences include a primary image sequence captured by a primary camera in the multi-camera array and one or more secondary image sequences captured by respective secondary cameras in the multi-camera array, wherein the primary camera captures live-action scenes with a user-specified composition and focus, and wherein each of the secondary cameras captures volumetric information corresponding to the live-action scenes from a different vantage point; generating, from at least the primary image sequence and one or more of the secondary image sequences, a respective depth map for each of a plurality of images of the primary image sequence; automatically determining, based on at least one of the depth maps and corresponding image or images of the primary image sequence, an object depicted in the primary image sequence; automatically separating the determined object onto a distinct layer associated with at least one of the depth maps and a distinct layer or layers associated with a corresponding image or images; modifying an appearance of the object in at least one of the distinct layer associated with the at least one of the depth maps and the distinct layer or layers associated with the corresponding image or images so as to change an appearance of the object in stereoscopic images; and generating, for each image in the plurality of images of the primary image sequence, based on at least the image and the corresponding depth map, at least two images of a stereoscopic image sequence.