Depth map generation method, related system and computer program product
AssigneeSTMicroelectronics S.r.l.
InventorMassimiliano Barone
Life of this patent
Estimating the term…
Abstract
A depth map is generated from at least a first and a second image. A plurality of reference pixels in the first image are selected and associated with respective pixels in the second image. A disparity between each reference pixel and the respective pixel in said second image is determined, and a depth value is determined as a function of the respective disparity. The plurality of reference pixels is selected based on detected contours in the first image.
Claims26
A method, comprising:
- detecting contours in a first image;
- determining at least one sampling step size based on the detected contours;
- selecting a plurality of reference pixels in said first image based on the detected contours and the at least one sampling step size;
- associating with each reference pixel a respective pixel in a second image;
- determining a disparity between each reference pixel and the respective associated pixel in said second image;
- and
- determining for each reference pixel a depth value as a function of the respective disparity.
A system, comprising:
- one or more memories;
- and
- one or more processing devices, which, in operation:
- detect contours in a first image;
- determine at least one sampling step size based on detected contours;
- select a plurality of reference pixels in said first image based on detected contours and determined sampling step sizes;
- associate reference pixels of the plurality with respective pixels in a second image;
- determine disparities between reference pixels and respective associated pixels in said second image;
- and
- determine reference pixel depth values based on the determined disparities.
A system, comprising:
- a contour detector to detect contours in a first image;
- a pixel sampler to sample the first image using step sizes based on detected contours;
- and
- a depth imager to associate pixels in a second image with the sampled pixels of the first image, determine disparities between associated pixels of the first and second images and determine depth values of the sampled pixels of the first image based on based the determined disparities.
A non-transitory computer-readable medium whose contents configure a digital image processing system to perform a method, the method comprising:
- detecting contours in a first image;
- determining at least one sampling step size based on the detected contours;
- selecting a plurality of reference pixels in said first image based on the detected contours and the determined sampling step sizes;
- associating reference pixels of the plurality with respective pixels in a second image;
- determining disparities between reference pixels and respective associated pixels in said second image;
- and
- determining reference pixel depth values based on the determined disparities.
Dependencies are inferred from the claim text.