Loading patent…
AssigneeMicrosoft Technology Licensing, LLC
InventorsAlvaro Collet Romea, Bao Zhang, Adam G. Kirk
Estimating the term…
The subject disclosure is directed towards a framework that is configured to allow different background-foreground segmentation modalities to contribute towards segmentation. In one aspect, pixels are processed based upon RGB background separation, chroma keying, IR background separation, current depth versus background depth and current depth versus threshold background depth modalities. Each modality may contribute as a factor that the framework combines to determine a probability as to whether a pixel is foreground or background. The probabilities are fed into a global segmentation framework to obtain a segmented image.
A system comprising, a foreground background segmentation framework, including a multimodal segmentation algorithm configured to accept contribution factors from different segmentation modalities and process the contribution factors to determine foreground versus background data for each element of an image that is useable to determine whether that element is a foreground or background element.
A method, comprising, processing a frame of image data and processing depth data computed from a corresponding depth-related image, including performing background subtraction on an element of the image data to obtain a background subtraction contribution factor for that element, determining one or more other depth-based contribution factors based upon the depth data associated with that element, computing a combined data term based at least in part upon a contribution from the background contribution factor and a contribution from each of the one or more other depth-based contribution factors, and using the data term in conjunction with other data terms as input to a global binary segmentation mechanism to obtain a segmented image.
One or more machine-readable storage media or logic having executable instructions, which when executed perform steps, comprising:
Dependencies are inferred from the claim text.