Virtual Masking Using Rigid Parametric Modeling
AssigneeSeiko Epson Corporation
InventorsJing Xiao, Kar-Han Tan, Anoop K. Bhattacharjya
Life of this patent
Estimating the term…
Abstract
Disclosed are systems and methods for masking at least a portion of one image with another image. Aspects of the present invention facilitate the placing of a virtual mask onto an item in an image even if that items moves in subsequent images, such as between different image frames in a video.
Claims20
A method for masking for a target item in an image, the method comprising the steps of:
- constructing a three-dimensional rigid parametric model of the target item;
- constructing a mapping between a mask image and a target image of the target item to obtain a mapped mask image;
- obtaining three-dimensional pose information of the target item in a captured image comprising pixels;
- warping the mapped mask image to the three-dimensional rigid parametric model, the warped mapped mask image comprising pixels;
- projecting at least some of the pixels in the warped mapped mask image to pixel locations in the captured image using the three-dimensional pose information of the target item in the captured image;
- and
- forming a composite image, wherein the target item in the captured image is at least partially masked by at least a portion of pixels projected from the warped mapped mask image.
A method of masking an item in an image comprising pixels wherein each pixel has a position and a color value, the method comprising the steps of:
- constructing a three-dimensional rigid parametric model of a target image of the item;
- constructing a mapping the target image and a mask image to obtain a mapped mask image;
- warping the mapped mask image to the three-dimensional rigid parametric model;
- identifying the item in a captured image;
- obtaining three-dimensional orientation and position information of the item in the captured image using the three-dimensional rigid parametric model;
- and
- generating a composite image by projecting at least some of the pixels of the warped mapped mask image to pixel locations in the captured image using the three-dimensional orientation and position information of the item in the captured image.
A system for masking a target item in an image, the system comprising:
- a model generator, coupled to receive information regarding a target image of the target item, that constructs a three-dimensional rigid parametric model of the target item;
- a mapper, coupled to receive information regarding the target image, that construct a mapping between a mask image comprising pixels and the target image to obtain a mapped mask image;
- a warper, coupled to receive information regarding the mapped mask image, that warps the mapped mask image to the three-dimensional rigid parametric model;
- a three-dimensional pose estimator, coupled to receive a captured image, that obtaining three-dimensional orientation and position information of the target item in the captured image;
- a projector, coupled to receive the three-dimensional orientation and position information, that projects at least some of the pixels of the warped mapped mask image to pixel locations in the captured image using the three-dimensional orientation and position information of the target item in the captured image to determine the pixel locations in the captured image;
- and
- an image blender, coupled to receive information regarding the projected pixels of the warped mapped mask image and regarding the captured image, that blends at least some of the pixels projected from the warped mapped mask image with corresponding pixels of the captured image to generate a composite image.
Dependencies are inferred from the claim text.