Back to results
US 8,139,153 B2Granted
Method for generating distances representative of the edge orientations in a video picture, corresponding device and use of the method for deinterlacing or format conversion
AssigneeThomson Licensing
InventorsJonathan Kervec, Pascal Bourdon, Nicolas Menard
Life of this patent
Estimating the term…
Abstract
The distances representative of the edge orientation at a current points of a video picture generated are particularly used for the deinterlacing or format conversion of video pictures. It can be implemented in a video coder or decoder. According to the invention, the distances representative of the edge orientation calculated are filtered with a conditional median filter applied to a window of p by q distances. This filter only takes into account the distances of points for which the local gradient module of the video components is greater than a threshold.
Claims9
A method, comprising:
- calculating, at a current point of a video picture and at a plurality of points of said video picture surrounding said current point forming a set of k points centered on said current point, distances representative of edge orientations at said points, k being a natural integer greater than one, the distances being calculated from local gradients of video components of said video picture at said points;
- and
- filtering, via a conditional median filter, the distance calculated for the current point on a set of distances, among the calculated distances for the points of the set of k points, respecting a predetermined criterion, the distances respecting said predetermined criterion being the distances of points whose module of local gradient of video components at said points is greater than a threshold.
A device, comprising:
- a calculation circuit operative to calculate, at a current point of a video picture and at a plurality of points of said video picture surrounding said current point forming a set of k points centered on said current point, distances representative of edge orientations at said points, k being a natural integer greater than one, the distances being calculated from local gradients of video components of said video picture at said points;
- and
- a conditional median filter operative to filter the distance calculated for said current point on a set of distances, among the calculated distances for the points of the set of k points, respecting a predetermined criterion, the distances respecting said predetermined criterion being the distances of points whose module of local gradient of video components at said points is greater than a threshold.
A method for de-interlacing a video picture, said method comprising:
- calculating, at a current point of said video picture and at a plurality of points of said video picture surrounding said current point forming a set of k points centered on said current point, distances representative of edge orientations at said points, k being a natural integer greater than one, the distances being calculated from local gradients of video components of said video picture at said points;
- filtering, via a conditional median filter, the distance calculated for said current point on a set of distances, among the calculated distances for the points of the set of k points, respecting a predetermined criterion, the distances respecting said predetermined criterion being the distances of points whose module of local gradient of video components at said points is greater than a threshold;
- and
- calculating values of pixels of a de-interlaced picture using the distance representative of the edge orientation calculated at said current point of the video picture.
1 more element
A method for converting an input video picture to an output video picture, said method comprising:
- calculating, at a current point of said input video picture and at a plurality of points of said input video picture surrounding said current point forming a set of k points centered on said current point, distances representative of edge orientations at said points, k being a natural integer greater than one, the distances being calculated from local gradients of video components of said input video picture at said points;
- filtering, via a conditional median filter, the distance calculated for said current point on a set of distances, among the calculated distances for the points of the set of k points, respecting a predetermined criterion, the distances respecting said predetermined criterion being the distances of points whose module of local gradient of video components at said points is greater than a threshold;
- and
- calculating values of pixels of the output video picture using the distance representative of the edge orientation calculated at said current point of the input video picture.
1 more element
Dependencies are inferred from the claim text.