SYSTEMS AND METHODS FOR IMAGE DEPTH MAP GENERATION

摘要

Systems and methods which provide generation of image depth maps which more accurately represent the local depth discontinuity within images through use of image global depth maps adapted based upon image global motion and/or localized depth analysis utilizing relative relationships of attributes across depth discontinuities in the image are disclosed. Embodiments utilize a full global depth map which is larger than or equal to the image being converted in order to accommodate image global motion, in generating an image global depth map. In operation according to embodiments, an image global depth map is identified within the full global depth map, such as based upon global motion within the image. Localized depth analysis, using pixel attribute relative relationships, is applied with respect to the image global depth map according to embodiments to generate an image depth map which more accurately reflects the local depth discontinuities within the image.

主权项

1. A method comprising:
receiving a mono-view image for generating an image depth map of the mono-view image, wherein the mono-view image is part of a plurality of images; tracking global motion of the mono-view image to provide image global motion information, wherein the global motion comprises general movement of the mono-view image which does not amount to a scene change within the motion video segment; generating an image global depth map using the image global motion information; and refining local depth values of the image global depth map using a relative relation of a plurality of color components for pixels of the mono-view image to provide the image depth map of the mono-view image.