Low complexity color smoothing of reconstructed point clouds

1. A decoding device for point cloud decoding, the decoding device comprising:

a communication interface configured to receive a bitstream; and

a processor operably coupled to the communication interface, wherein the processor is configured to:

decode the bitstream into multiple frames that include patches corresponding to respective clusters of points from a 3D point cloud;

generate a grid that includes a plurality of 3D cells, wherein the 3D point cloud is within the grid;

identify a first cell, of the plurality of 3D cells, that includes a query point corresponding to a pixel on a boundary of one of the patches;

identify a luminance value of the first cell; and

perform color smoothing on the query point based on comparison of the luminance value of the first cell to a threshold.

2. The decoding device of claim 1 , wherein to identify the luminance value of the first cell, the processor is configured to:

identify luminance values of the points within the first cell;

identify a mean value and a median value based on the luminance values of the points within the first cell; and

identify the luminance value of the first cell based on a difference between the mean value and the median value.

3. The decoding device of claim 1 , wherein to smooth the query point, the processor is configured to:

identify neighboring cells of the first cell, based on a position of the query point within the first cell; and

generate a new color for the query point based on the neighboring cells and the first cell.

4. The decoding device of claim 3 , wherein the processor is further configured to:

identify luminance values of the points within a first neighboring cell of the neighboring cells;

identify a mean value and a median value based on the luminance values of the points within the first neighboring cell;

exclude the first neighboring cell as one of the neighboring cells when a difference between the mean value and the median value of the first neighboring cell is greater than the threshold; and

generate the new color for the query point based on the first cell and the neighboring cells that remain.

5. The decoding device of claim 3 , wherein the processor is further configured to:

identify a luminance value of a first neighboring cell of the neighboring cells;

exclude the first neighboring cells as one of the neighboring cells when a difference between the luminance value of the first cell and the luminance value of the first neighboring cells is greater than a second threshold; and

generate the new color for the query point based on the first cell and the neighboring cells that remain.

6. The decoding device of claim 3 , wherein generate the new color, the processor is configured to:

identify a color centroid of each of the neighboring cells and the first cell; and

generate the new color based on a weighted average of the color centroid of each of the neighboring cells and the first cell.

7. The decoding device of claim 6 , wherein the processor is configured to apply a tri-linear filter to the neighboring cells and the first cell to generate the new color.

8. The decoding device of claim 3 , wherein the processor is further configured to:

identify a color of the query point within the first cell;

determine a distance between the new color and the color of the query point; and

replace the color of the query point with the new color, when the distance is greater than another threshold.

9. The decoding device of claim 1 , wherein, to smooth the query point, the processor is configured to:

identify a color centroid of the first cell,

replace a color of the query point with the color centroid of the first cell.

10. The decoding device of claim 1 , wherein the plurality of 3D cells within the grid do not overlap.

11. A method for point cloud decoding comprising:

receiving a bitstream;

decoding the bitstream into multiple frames that include patches corresponding to respective clusters of points from a 3D point cloud;

generating a grid that includes a plurality of 3D cells, wherein the 3D point cloud is within the grid;

identifying a first cell, of the plurality of 3D cells, that includes a query point corresponding to a pixel on a boundary of one of the patches;

identifying a luminance value of the first cell; and

performing color smoothing on the query point based on comparison of the luminance value of the first cell to a threshold.

12. The method of claim 11 , wherein identifying the luminance value of the first cell comprises:

identifying luminance values of the points within the first cell;

identifying a mean value and a median value based on the luminance values of the points within the first cell; and

identifying the luminance value of the first cell based on a difference between the mean value and the median value.

13. The method of claim 11 , wherein smoothing the query point comprises:

identifying neighboring cells of the first cell, based on a position of the query point within the first cell; and

generating a new color for the query point based on the neighboring cells and the first cell.

14. The method of claim 13 , further comprising:

identifying luminance values of the points within a first neighboring cell of the neighboring cells;

identifying a mean value and a median value based on the luminance values of the points within the first neighboring cell;

excluding the first neighboring cell as one of the neighboring cells when a difference between the mean value and the median value of the first neighboring cell is greater than the threshold; and

generating the new color for the query point based on the first cell and the neighboring cells that remain.

15. The method of claim 13 , further comprising:

identifying a luminance value of a first neighboring cell of the neighboring cells;

excluding the first neighboring cells as one of the neighboring cells when a difference between the luminance value of the first cell and the luminance value of the first neighboring cells is greater than a second threshold; and

generating the new color for the query point based on the first cell and the neighboring cells that remain.

16. The method of claim 13 , wherein generating the new color comprises:

identifying a color centroid of each of the neighboring cells and the first cell; and

generating the new color based on a weighted average of the color centroid of each of the neighboring cells and the first cell.

17. The method of claim 16 , further comprises applying apply a tri-linear filter to the neighboring cells and the first cell to generate the new color.

18. The method of claim 13 , further comprising:

identifying a color of the query point within the first cell;

determining a distance between the new color and the color of the query point; and

replacing the color of the query point with the new color, when the distance is greater than another threshold.

19. The method of claim 11 , wherein smoothing the query point comprises:

identifying a color centroid of the first cell,

replacing a color of the query point with the color centroid of the first cell.

20. The method of claim 11 , wherein the plurality of 3D cells within the grid do not overlap.