Multi-channel depth estimation using census transforms

1. A method comprising:

capturing, via a first camera, a first image that includes a plurality of light channels;

capturing, via a second camera, a second image that includes the plurality of light channels;

selecting a scan direction from a plurality of scan directions in association with the first image and the second image;

along each of a plurality of scanlines in the selected scan direction, comparing pixels from the first image to pixels from the second image based on calculating a census transform for each pixel in the first image and a census transform for each pixel in the second image;

determining a stereo correspondence between the pixels in the first image and the pixels in the second image based on the comparing; and

generating a depth map based on the stereo correspondence.

2. The method of claim 1 , wherein selecting the scan direction from the plurality of scan directions comprises selecting an angle made by the scan direction to a horizontal direction.

3. The method of claim 2 , wherein selecting an angle made by the scan direction to the horizontal direction further comprises selecting at least one of:

an angle of 0° from the horizontal direction;

an angle of 90° from the horizontal direction; and

an angle between 0° and 90° from the horizontal direction.

4. The method of claim 1 , wherein the plurality of scanlines in the selected scan direction are parallel to each other.

5. The method of claim 1 , wherein the plurality of scanlines in the selected scan direction cover all pixels in at least one of: the first image and the second image.

6. The method of claim 1 , wherein calculating a census transform for each pixel in the first image and a census transform for each pixel in the second image comprises:

in a first light channel of the plurality of light channels, calculating a census transform for each pixel of the first image to form a first set of census transforms and a census transform for each pixel of the second image to form a second set of census transforms; and

in a second light channel of the plurality of light channels, calculating a census transform for each pixel of the first image to form a third set of census transforms and a census transform for each pixel of the second image to form a fourth set of census transforms.

7. The method of claim 6 , wherein comparing pixels from the first image to pixels from the second image based on calculating the census transform for each pixel in the first image and the census transform for each pixel in the second image comprises:

for a first pixel in the first image and a second pixel in the second image:

calculating a first Hamming distance for the first light channel between the census transform of the first pixel and the census transform of the second pixel;

calculating a second Hamming distance for the second light channel between the census transform of the first pixel and the census transform of the second pixel; and

summing the Hamming distances;

wherein the stereo correspondence is determined in part by minimizing the sum of the Hamming distances.

8. The method of claim 6 , wherein generating the depth map further comprises:

for each selected scan direction of the plurality of scan directions:

along each scanline of the plurality of scanlines in the scan direction, comparing the first set of census transforms to the second set of census transforms and the third set of census transforms to the fourth set of census transforms;

wherein the stereo correspondence between the pixels of the first image and the pixels of the second image is based on the comparisons in each scan direction.

9. The method of claim 1 , wherein at least two light channels, of the plurality of light channels, include at least two different color channels.

10. The method of claim 1 , wherein at least two light channels, of the plurality of light channels, include a color channel and an infrared channel.

11. A non-transitory computer-readable storage medium storing instructions that, when executed by a processor, cause the processor to perform steps of:

capturing, via a first camera, a first image that includes a plurality of light channels;

capturing, via a second camera, a second image that includes the plurality of light channels;

selecting a scan direction from a plurality of scan directions in association with the first image and the second image;

along each of a plurality of scanlines in the selected scan direction, comparing pixels from the first image to pixels from the second image based on calculating a census transform for each pixel in the first image and a census transform for each pixel in the second image;

determining a stereo correspondence between the pixels in the first image and the pixels in the second image based on the comparing; and

generating a depth map based on the stereo correspondence.

12. The storage medium of claim 11 , wherein selecting the scan direction from the plurality of scan directions comprises selecting an angle made by the scan direction to a horizontal direction.

13. The storage medium of claim 12 , wherein selecting an angle made by the scan direction to the horizontal direction further comprises selecting at least one of:

an angle of 0° from the horizontal direction;

an angle of 90° from the horizontal direction; and

an angle between 0° and 90° from the horizontal direction.

14. The storage medium of claim 11 , wherein the plurality of scanlines in the selected scan direction are parallel to each other.

15. The storage medium of claim 11 , wherein the plurality of scanlines in the selected scan direction cover all pixels in at least one of: the first image and the second image.

16. A depth estimation system comprising:

a first camera configured to capture images in a plurality of light channels;

a second camera configured to capture images in the plurality of light channels; and

a controller configured to perform operations comprising:

selecting a scan direction from a plurality of scan directions in association with a first image captured by the first camera and a second image captured by the second camera;

along each of a plurality of scanlines in the selected scan direction, comparing pixels from the first image to pixels from the second image based on calculating a census transform for each pixel in the first image and a census transform for each pixel in the second image;

determining a stereo correspondence between the pixels in the first image and the pixels in the second image based on the comparing; and

generating a depth map based on the stereo correspondence.

17. The depth estimation system of claim 16 , wherein selecting the scan direction from the plurality of scan directions comprises selecting an angle made by the scan direction to a horizontal direction.

18. The depth estimation system of claim 17 , wherein selecting an angle made by the scan direction to the horizontal direction further comprises selecting at least one of:

an angle of 0° from the horizontal direction;

an angle of 90° from the horizontal direction; and

an angle between 0° and 90° from the horizontal direction.

19. The depth estimation system of claim 16 , wherein the plurality of scanlines in the selected scan direction are parallel to each other.

20. The depth estimation system of claim 16 , wherein the plurality of scanlines in the selected scan direction cover all pixels in at least one of: the first image and the second image.