Predictive coding for point cloud compression

One or more non-transitory, computer-readable storage media, storing program instructions that when executed on or across one or more computing devices cause the one or more computing devices to:

• receive encoded node information for nodes corresponding to a plurality of points representing a three-dimensional (3D) scene, wherein one or more child nodes and one or more ancestor nodes of the one or more child nodes correspond to respective points of the 3D scene;
• decode the encoded node information, wherein, in decoding the encoded node information, the program instructions cause the one or more computing devices to further:
• decode respective indicated child nodes according to respective prediction techniques indicated in the node information, wherein values of at least one of the respective points of the respective child nodes were predicted according to the respective prediction techniques;
• and
• store or render the plurality of points of the 3D scene decoded from the node information.

A system, comprising:

• a memory storing program instructions, and
• one or more processors configured to execute the program instructions to:
• generate a prediction tree comprising a plurality of nodes that correspond to a plurality of points that represent a three-dimensional (3D) scene captured using one or more sensors, wherein respective ones of the points captured using the one or more sensors comprise spatial information for the respective points, and wherein, in generating the prediction tree, the program instructions cause the one or more computing devices to:
• select individual ones of the plurality of points to include in respective ones of the plurality of nodes in the prediction tree;
• determine respective predicted node values, for the individual ones of the points, determined using respective prediction techniques applied to one or more ancestor nodes of one or more nodes of the prediction tree representing the individual ones of the points, wherein the one or more ancestor nodes correspond to one or more other nodes in the prediction tree for which corresponding node values are known or have been predicted for one or more other points of the 3D scene;
• encode node information of the prediction tree, wherein the encoded node information comprises, for a given node, an indicator of the respective prediction technique applied to determine the respective predicted node value for the given node;
• and
• send or store the encoded node information.