Occluding contour detection and storage for digital photography

摘要

Described are techniques used in occluding contour detection. Cues obtained from a static image are combined with cues obtained based on motion disparity of objects in the image to determine occluding image contours. Two images of a same scene are obtained. Local pixel-wise flow fields are obtained using the Lucas-Kanade technique. A warped image is determined using the second image and the flow field. A disparity map is determined by comparing the warped image to the original first image. Static image cues are obtained from the original image using the Berkeley technique. The static image cues are combined with the motion disparity cues indicated by the disparity map producing a product image map. Occluding contours are determined by processing the product image map with the JetStream tracking technique. The occluding contours may be used with the original image for subsequent processing.