Depth acquisition device and depth acquisition method

1. A depth acquisition device, comprising:

a memory; and

a processor, wherein

the processor performs:

acquiring timing information indicating a timing at which a light source irradiates a subject with infrared light;

acquiring an infrared light image stored in the memory, the infrared light image being generated by imaging a scene including the subject with the infrared light according to the timing indicated by the timing information;

acquiring a visible light image stored in the memory, the visible light image being generated by imaging a substantially same scene as the scene of the infrared light image, with visible light from a substantially same viewpoint as a viewpoint of the imaging the infrared light image at a substantially same time as an imaging time of imaging the infrared light image;

detecting a flare region from the infrared light image; and

estimating a depth of the flare region based on the infrared light image, the visible light image, and the flare region,

wherein

in the estimating of the depth of the flare region, the processor performs:

estimating depth information indicating a depth at each position in the infrared light image; and

correcting a depth at each position in the flare region which is indicated in the depth information, by inputting the infrared light image, the visible light image, the flare region, and the depth information into a learning model.

2. The depth acquisition device according to claim 1 , wherein

in the estimating of the depth of the flare region, the processor performs:

estimating first depth information indicating a depth at each position in the infrared light image;

estimating second depth information indicating a corrected depth at each position in the flare region, the corrected depth being obtained by correcting, based on the visible light image, a depth at each position in the flare region which is indicated in the first depth information; and

generating third depth information indicating (i) a depth at each position in a region other than the flare region in the infrared light image which is indicated in the first depth information and (ii) a depth at each position in the flare region in the infrared light image which is indicated in the second depth information.

3. The depth acquisition device according to claim 1 , wherein

in the detecting of the flare region, the processor performs

detecting, as the flare region, a region having a luminance not less than a first threshold in the infrared light image.

4. The depth acquisition device according to claim 1 , wherein

in the detecting of the flared region, the processor performs

detecting, as the flared region, a region having a luminance not less than a first threshold and satisfying a predetermined condition in the infrared light image, and

the predetermined condition is that a correlation value between (i) an image feature amount in a first region of the infrared light image and (ii) an image feature amount in a second region of the visible light image is less than a second threshold, the second region corresponding to the first region.

5. The depth acquisition device according to claim 4 , wherein

the image feature amount in the first region of the infrared light image is an edge included in the first region, and

the image feature amount in the second region of the visible light image is an edge included in the second region.

6. The depth acquisition device according to claim 4 , wherein

the image feature amount in the first region of the infrared light image is a luminance in the first region, and

the image feature amount in the second region of the visible light image is a luminance in the second region.

7. The depth acquisition device according to claim 1 , wherein

in the detecting of the flared region, the processor performs for each of one or more higher luminance regions each having a luminance not less than a first threshold in the infrared light image:

(i) generating a first converted image by executing Census transform on an image in the each of the one or more higher luminance regions;

(ii) generating a second converted image by executing Census transform on an image in a region of the visible light image, the region corresponding to the each of the one or more higher luminance regions; and

detecting, as the flare region, a higher luminance region having a Humming distance greater than a third threshold from the one or more higher luminance regions, the Humming distance being a Humming distance between the first converted image and the second converted image.

8. A depth acquisition method, comprising:

acquiring timing information indicating a timing at which a light source irradiates a subject with infrared light;

acquiring an infrared light image stored in a memory, the infrared light image being generated by imaging a scene including the subject with the infrared light according to the timing indicated by the timing information;

acquiring a visible light image stored in the memory, the visible light image being generated by imaging a substantially same scene as the scene of the infrared light image, with visible light from a substantially same viewpoint as a viewpoint of the imaging the infrared light image at a substantially same time as an imaging time of imaging the infrared light image;

detecting a flare region from the infrared light image; and

estimating a depth of the flare region based on the infrared light image, the visible light image, and the flare region,

wherein

in the estimating of the depth of the flare region, the processor performs:

estimating depth information indicating a depth at each position in the infrared light image; and

correcting a depth at each position in the flare region which is indicated in the depth information, by inputting the infrared light image, the visible light image, the flare region, and the depth information into a learning model.