Methods and devices for encoding and reconstructing a point cloud

    公开(公告)号:US11627339B2

    公开(公告)日:2023-04-11

    申请号:US16616491

    申请日:2018-04-12

    Abstract: This method comprises: —accessing (2) a point cloud (PC) comprising a plurality of points defined by attributes, said attributes including a spatial position of a point in a 3D space and at least one feature of the point;—segmenting (2) the point cloud into one or more clusters (Ci) of points on the basis of the attributes of the points; and for at least one cluster (Ci):—constructing (4) a similarity graph having a plurality of vertices and at least one edge, the similarity graph representing a similarity among neighboring points of the cluster (Ci) in terms of the attributes, the plurality of vertices including vertices Pi and Pj corresponding to points of the cluster (Ci);—assigning one or more weights wi,j to one or more edges connecting vertices Pi and Pj of the graph;—computing (6) a transform using the one or more assigned weights, said transform being characterized by coefficients; and—quantizing (8) and encoding (10) the transform coefficients.

    Method and apparatus for reconstructing a point cloud representing a 3D object

    公开(公告)号:US11508041B2

    公开(公告)日:2022-11-22

    申请号:US16753773

    申请日:2018-10-03

    Abstract: The present disclosure concerns a method for reconstructing a point cloud representing a 3D object from an inverse-projected point cloud obtained by inverse-projecting at least one depth image of an original point cloud, said at least one depth image being obtained by projecting points of the original point cloud onto at least one surface, said method comprising the steps of detecting at least one missing part in the inverse-projected point cloud, and completing said at least one missing part based on points in the neighborhood of said at least one missing part.

    Processing a point cloud
    9.
    发明授权

    公开(公告)号:US11949889B2

    公开(公告)日:2024-04-02

    申请号:US16962646

    申请日:2019-01-21

    CPC classification number: H04N19/182 H04N19/80

    Abstract: At least one embodiment relates to a method for padding a first depth image representative of depth values of nearer points of a point cloud frame and a second depth image representative of depth values of farther points of a point cloud frame. The method also comprises encoding a video stream including a time-interleaving of said encoded first and second images. There is also provided a method comprising decoding a video stream to provide a first depth image representative of depth values of nearer points of a point cloud frame and a second depth image representative of depth values of farther points of a point cloud frame; and filtering pixel values of the second depth image by using pixel values of the first depth image.

    Method and device for up-sampling a point cloud

    公开(公告)号:US11455748B2

    公开(公告)日:2022-09-27

    申请号:US16753782

    申请日:2018-10-02

    Abstract: This method for up-sampling a point cloud representing a 3D object, comprises: —detecting (2) points belonging to at least one under-sampled region of the point cloud on the basis of at least one desirable sampling rate (K); —obtaining (12), for each detected point, an associated tangent plane; —inserting (14) in the point cloud at least one neighboring point of each detected point if a distance between the neighboring point and the tangent plane associated with the detected point is less than a first threshold.

Patent Agency Ranking