Video segmentation from an uncalibrated camera array

[0015]Aspects presented herein provide an approach for image segmentation from an uncalibrated camera array. In contrast to traditional green screens and chroma keying, a segmentation application obtains depth information for pixels of an image and performs a depth-based segmentation based on a “virtual” placement of a green screen after the image is captured. In one aspect, the segmentation application computes a pseudo depth map for each frame of a video sequence recorded with a camera array, based on dense correspondences between cameras in the array. As used herein, a pseudo depth generally refers to a depth ordering of pixels indicating a relative depth of each pixel with respect to other pixels in an image, without reconstructing actual depths of the pixels. Computing pseudo depths does not require geometric or photometric camera calibration (i.e., the camera array may be uncalibrated), and there is no metric meaning behind pseudo depths, i.e., the pseudo depths do not actually r