Apparatus and method for noise reduction in depth images during object segmentation

1. A image noise reduction method comprising:

capturing an image, including capturing depth information of at least one object in the image;

generating a depth map of the image including the depth information of the at least one object;

generating segmentation information corresponding to the image by over-segmenting the depth map;

determining at least one segment to be noise based on the segmentation information; and

deleting the at least one segment determined to be noise.

2. The method of claim 1 , wherein the at least one segment determined to be noise comprises a segment having a size under a threshold.

3. The method of claim 1 , further comprising:

virtually rotating a camera viewpoint;

repeating the over-segmenting of the depth map;

determining another segment to be noise; and

deleting the other segment determined to be noise, according to the rotated viewpoint,

wherein the captured image is captured from a different viewpoint than the rotated viewpoint.

4. The method of claim 1 , further comprising:

voxelizing a point cloud; and

deleting at least one voxel determined to be noise.

5. The method of claim 4 , wherein the voxel determined to be noise comprises a voxel with fewer than a threshold number of votes.

6. The method of claim 4 , wherein the voxel determined to be noise comprises a voxel with fewer than a threshold number of neighbors.

7. The method of claim 1 , wherein the over-segmenting is based on a Connected Components (CC) algorithm.

8. The method of claim 7 , wherein the CC algorithm comprises a depth smoothness threshold of at least a predetermined size.

9. The method of claim 8 , wherein the at least one segment determined to be noise comprises a segment having a size under the depth smoothness threshold.

10. An apparatus for image noise reduction, the apparatus comprising:

a camera configured to capture an image;

a depth sensor configured to determine depth information of at least one object in the image; and

at least one processor configured to:

generate a depth map of the image according to the depth information of the at least one object;

generate segmentation information corresponding to the image, by over-segmenting the depth map;

determine at least one segment to be noise based on the segmentation information; and

delete the at least one segment determined to be noise.

11. The apparatus of claim 10 , wherein the at least one segment determined to be noise comprises a segment having a size under a threshold.

12. The apparatus of claim 10 , wherein the at least one processor is further configured to:

virtually rotate a camera viewpoint;

repeat the over-segmenting of the depth map;

determine another segment to be noise; and

delete the other segment determined to be noise, according to the rotated viewpoint,

wherein the captured image is captured from a different viewpoint than the rotated viewpoint.

13. The apparatus of claim 10 , wherein the at least one processor is further configured to:

voxelize a point cloud; and

delete at least one voxel determined to be noise.

14. The apparatus of claim 13 , wherein the voxel determined to be noise comprises a voxel with fewer than a threshold number of votes.

15. The apparatus of claim 13 , wherein the voxel determined to be noise comprises a voxel with fewer than a threshold number of neighbors.

16. The apparatus of claim 10 , wherein the over-segmenting is based on a Connected Components (CC) algorithm.

17. The apparatus of claim 16 , wherein the CC algorithm comprises a depth smoothness threshold of at least a predetermined size.

18. The apparatus of claim 17 , wherein the at least one segment determined to be noise comprises a segment having a size under the depth smoothness threshold.