Progressive lossless point cloud attribute coding

1 . A method of decoding point cloud data, comprising:

obtaining upsampled attributes of a previous octree level of the point cloud data;

predicting attributes of a current voxel group of a current octree level of the point cloud data, based on attributes of already decoded voxel groups in the current octree level and the upsampled attributes of the previous octree level;

obtaining a probability distribution of the current voxel group, based on the predicted attributes of the current voxel group; and

decoding the current voxel group from an input bitstream, based on the probability distribution of the current voxel group.

2 . The method of claim 1 , wherein the predicting attributes of a current voxel group is based on a sum of attribute values of already decoded voxel groups.

3 . The method of claim 1 , wherein the predicting attributes of a current voxel group is based on that an average of remaining attribute values to be predicted and already decoded voxel groups at the current octree level being same as an attribute value of a corresponding parent voxel at a previous octree level.

4 . The method of claim 1 , wherein attributes of the current octree level are predicted sequentially from one voxel group to next voxel group.

5 . The method of claim 1 , wherein the predicting attributes of a current voxel group is based on a neural network.

6 . A method of encoding point cloud data, comprising:

obtaining upsampled attributes of a previous octree level of the point cloud data;

predicting attributes of a current voxel group of a current octree level of the point cloud data, based on attributes of already encoded voxel groups in the current octree level and the upsampled attributes of the previous octree level;

obtaining a probability distribution of the current voxel group, based on the predicted attributes of the current voxel group; and

encoding the current voxel group based on the probability distribution of the current voxel group.

7 . The method of claim 6 , wherein the predicting attributes of a current voxel group is based on a sum of attribute values of already encoded voxel groups.

8 . The method of claim 6 , wherein the predicting attributes of a current voxel group is based on that an average of remaining attribute values to be predicted and already decoded voxel groups at the current octree level being same as an attribute value of a corresponding parent voxel at a previous octree level.

9 . The method of claim 6 , wherein attributes of the current octree level are predicted sequentially from one voxel group to next voxel group.

10 . The method of claim 6 , wherein the predicting attributes of a current voxel group is based on a neural network.

11 . An apparatus for decoding point cloud data, comprising one or more processors and at least one memory coupled to the one or more processors, wherein the one or more processors are configured to:

obtain upsampled attributes of a previous octree level of the point cloud data;

predict attributes of a current voxel group of a current octree level of the point cloud data, based on attributes of already decoded voxel groups in the current octree level and the upsampled attributes of the previous octree level;

obtain a probability distribution of the current voxel group, based on the predicted attributes of the current voxel group; and

decode the current voxel group from an input bitstream, based on the probability distribution of the current voxel group.

12 . The apparatus of claim 11 , wherein the predicting attributes of a current voxel group is based on a sum of attribute values of already decoded voxel groups.

13 . The apparatus of claim 11 , wherein the predicting attributes of a current voxel group is based on that an average of remaining attribute values to be predicted and already decoded voxel groups at the current octree level being same as an attribute value of a corresponding parent voxel at a previous octree level.

14 . The apparatus of claim 11 , wherein attributes of the current octree level are predicted sequentially from one voxel group to next voxel group.

15 . The apparatus of claim 11 , wherein the predicting attributes of a current voxel group is based on a neural network.

16 . An apparatus for encoding point cloud data, comprising one or more processors and at least one memory coupled to the one or more processors, wherein the one or more processors are configured to:

obtain upsampled attributes of a previous octree level of the point cloud data;

predict attributes of a current voxel group of a current octree level of the point cloud data, based on attributes of already encoded voxel groups in the current octree level and the upsampled attributes of the previous octree level;

obtain a probability distribution of the current voxel group, based on the predicted attributes of the current voxel group; and

encode the current voxel group based on the probability distribution of the current voxel group.

17 . The apparatus of claim 16 , wherein the predicting attributes of a current voxel group is based on a sum of attribute values of already encoded voxel groups.

18 . The apparatus of claim 16 , wherein the predicting attributes of a current voxel group is based on that an average of remaining attribute values to be predicted and already decoded voxel groups at the current octree level being same as an attribute value of a corresponding parent voxel at a previous octree level.

19 . The apparatus of claim 16 , wherein attributes of the current octree level are predicted sequentially from one voxel group to next voxel group.

20 . The apparatus of claim 16 , wherein the predicting attributes of a current voxel group is based on a neural network.