Method and device for predictive encoding/decoding of a point cloud
AssigneeInterDigital VC Holdings, Inc.
InventorsSebastien Lasserre, Saurabh Puri, Kangying Cai, Julien Ricard, Celine Guede
Life of this patent
Estimating the term…
Abstract
This method for inter-predictive encoding of a time-varying 3D point cloud including a series of successive frames divided in 3D blocks into at least one bitstream comprises encoding (20) 3D motion information including a geometric transformation comprising rotation information representative of a rotation transformation and translation information representative of a translation transformation, wherein the translation information comprises a vector ΔT representing an estimation error of the translation transformation.
Claims18
A predictive video encoding method comprising:
- obtaining 3D motion information comprising a geometric transformation, wherein the 3D motion information comprises a vector pointing to a matched 3D block, wherein the geometric transformation comprises a rotation transformation and a translation transformation, and wherein the rotation transformation is represented by a rotation matrix R;
- obtaining information representative of the rotation transformation;
- obtaining an estimation of the translation transformation based on the information representative of the rotation transformation, wherein the estimation of the translation transformation is obtained as a vector T* from a reconstructed rotation matrix Rrec;
- determining an estimation error ΔT between the translation transformation and the estimation;
- and
- encoding the estimation error ΔT.
A predictive video coding device comprising a processor configured to at least:
- obtain 3D motion information comprising a geometric transformation, wherein the 3D motion information comprises a vector pointing to a matched 3D block, wherein the geometric transformation comprises a rotation transformation and a translation transformation, and wherein the rotation transformation is represented by a rotation matrix R;
- obtain information representative of the rotation transformation;
- obtain an estimation of the translation transformation based on the information representative of the rotation transformation, wherein the estimation of the translation transformation is obtained as a vector T* from a reconstructed rotation matrix Rrec;
- determine an estimation error ΔT between the translation transformation and the estimation;
- and
- encode the estimation error ΔT.
A predictive video decoding method, comprising:
- obtaining rotation information for predicting a current target block of a point cloud;
- determining an estimate of a translation transformation, wherein the estimate of the translation transformation is determined using at least the rotation information and mean position of reconstructed points associated with a reference block corresponding to the current target block;
- obtaining an estimation error ΔT;
- determining translation information for predicting the current target block from the estimate of the translation transformation and the estimation error ΔT;
- and
- predicting the current target block of the point cloud, wherein the current target block is predicted using the rotation information and the translation information.
A predictive video decoding device, the predictive video decoding device comprising a processor configured to:
- obtain rotation information for predicting a current target block of a point cloud;
- determine an estimate of a translation transformation using at least the rotation information and mean position of reconstructed points associated with a reference block corresponding to the current target block;
- obtain an estimation error ΔT;
- determine translation information for predicting the current target block from the estimate of the translation transformation and the estimation error ΔT;
- and
- predict the current target block of the point cloud using the rotation information and the translation information.
Dependencies are inferred from the claim text.