Method and apparatus for point cloud compression

A method for point cloud compression, comprising:

• compressing, by a processor using a video coding algorithm, a two-dimensional geometry image associated with a point cloud;
• reconstructing, by the processor, a geometry reconstructed cloud according to the compressed geometry image of the point cloud;
• selecting, using an algorithm for selecting points for selective smoothing, at least one geometry sample inside a patch of the geometry reconstructed cloud;
• applying, by the processor, a filter to the selected at least one geometry sample inside the patch of the geometry reconstructed cloud to generate a smoothed geometry reconstructed cloud;
• generating, by the processor, a texture image for the point cloud based on the smoothed geometry reconstructed cloud such that the compressed geometry image and the texture image are used to generate a coded bitstream representing the point cloud;
• and
• generating prediction information of the point cloud of the coded bitstream, the prediction information indicating that the selective smoothing inside patches of the geometry reconstructed cloud is applied and indicating the algorithm for selecting the points for the selective smoothing.

An apparatus for point cloud decompression, comprising:

• processing circuitry configured to:
• decode, using a video coding algorithm, prediction information of a point cloud from a coded bitstream, the prediction information indicating that selective smoothing inside patches of a geometry reconstructed cloud is applied and indicating an algorithm for selecting points for the selective smoothing;
• reconstruct the geometry reconstructed cloud according to a two-dimensional geometry image of the point cloud, the geometry image being decoded from the coded bitstream;
• in response to a determination that the prediction information indicates that the selective smoothing inside the patches of the geometry reconstructed cloud is applied,
• select at least one geometry sample inside a patch of the geometry reconstructed cloud, the selecting based on the algorithm indicated in the prediction information;
• apply a filter to the selected at least one geometry sample inside the patch of the geometry reconstructed cloud to generate a smoothed geometry reconstructed cloud;
• and
• reconstruct points of the point cloud based on the smoothed geometry reconstructed cloud such that the reconstructed points of the point cloud represent a three-dimensional space based on the coded bitstream.