Method for filtering spurious pixels in a depth-map
AssigneeInterDigital CE Patent Holdings, SAS
InventorsBenoit VANDAME, Neus SABATER, Guillaume BOISSON
Life of this patent
Estimating the term…
Abstract
The disclosure concerns a method for filtering spurious pixels in a depth-map. It is important to filter a depth-map to remove the spurious pixels. One issue when filtering spurious pixels, is to remove them while preserving the boundaries of the real details of the depth-map. A known approach to filter spurious pixels is to use a bi-lateral filter. However, the bi-lateral filter is not efficient to remove the spurious pixels. The proposed solution enables to remove spurious pixels in a depth-map while preserving the sharp edges of the depth-map and smoothing faint details or noise.
Claims12
A computer implemented method for filtering pixels in a depth-map, the method comprising, for a pixel of the depth map called current pixel:
- determining at least one weight for at least one pixel, called neighboring pixel, belonging to a group of pixels of said depth-map centered on the current pixel, said weight being a function of a geometrical distance between the current pixel and said at least one neighboring pixel as well as of an intensity of a pixel adjacent to the current pixel and an intensity of said at least one neighboring pixel,
- determining a weight function corresponding to at least one adjacent pixel, the coefficients of which being said determined weight,
- filtering said current pixel by applying a convolution filter corresponding to a weight function maximizing a sum of all its coefficients.
A device capable of filtering pixels in a depth-map, said device comprising at least one hardware processor configured to, for a pixel of the depth map called current pixel:
- determine at least one weight for at least one pixel, called neighboring pixel, belonging to a group of pixels of said depth-map centered on the current pixel, said weight being a function of a geometrical distance between the current pixel and said at least one neighboring pixel as well as of an intensity of a pixel adjacent to the current pixel and an intensity of said at least one neighboring pixel,
- determine a weight function corresponding to at least one adjacent pixel, the coefficients of which being said determined weight,
- filter said current pixel by applying the convolution filter corresponding to a weight function maximizing a sum of all its coefficients.
Dependencies are inferred from the claim text.