Motion estimation device and motion estimation method using motion estimation device

1. A motion estimation device, comprising:

a first receiving circuit configured to receive a first input frame; and

a calculation circuit configured to perform motion estimation processing based on the first input frame, wherein

in the motion estimation processing, the calculation circuit estimates a distance for each pixel of the first input frame and estimates a reliability of the distance for each pixel of the first input frame based on pixel information of the first input frame, and

in the motion estimation processing, the calculation circuit is configured to:

detect a first feature point in the first input frame, and

select the first feature point as a second feature point if the estimated reliability of the distance for the pixel corresponding to first feature point is greater than or equal to a first reference value.

2. The motion estimation device according to claim 1 , wherein, in the motion estimation processing, the calculation circuit is configured to:

compare the second feature point to a third feature point in a key frame registered before the first input frame,

select the second feature point as a fourth feature point if the second feature point and the third feature point do not match, and

detect the fourth feature point as a moving point in the first input frame if the fourth feature point is in a predetermined region of the first input frame.

3. The motion estimation device according to claim 2 , wherein, in the motion estimation processing, the calculation circuit is configured to:

detect the fourth feature point as a non-moving point in the first input frame if the fourth feature point is not in the predetermined region of the first input frame,

detect the second feature point as a non-moving point in the first input frame if the second feature point and the third feature point match, and

calculate a distance to the non-moving point.

4. The motion estimation device according to claim 3 , wherein the calculation circuit is further configured to:

compare the estimated distance for each pixel of the first input frame to the calculated distance of the non-moving point.

5. The motion estimation device according to claim 1 , wherein the calculation circuit estimates the distance for each pixel of the first input frame and estimates the reliability of the distance for each pixel of the first input frame based on the pixel information of the first input frame and machine learning data.

6. The motion estimation device according to claim 5 , wherein

the first input frame is an image captured by a camera,

the pixel information of each pixel of the first input frame is a blur shape for each pixel, and

the machine learning data includes blur shapes corresponding to various distances in the image.

7. The motion estimation device according to claim 5 , wherein

the first input frame is a full-color image,

the pixel information of the first input frame is a pixel value, and

the machine learning data includes pixel values corresponding to various distances in the full-color image.

8. The motion estimation device according to claim 1 , wherein the distance for each pixel of the first input frame is a distance between a subject and a camera that provides the first input frame to the first receiving circuit.

9. The motion estimation device according claim 1 , further comprising:

a second receiving circuit configured to receive a second input frame, wherein

the calculation circuit, in the motion estimation processing, is further configured to:

estimate the distance for each pixel of the first input frame and estimate the reliability of the estimated distance for each pixel of the first input frame based on pixel information of the second input frame in addition to the pixel information of the first input frame.

10. The motion estimation device according to claim 1 , wherein, in the motion estimation processing, the calculation circuit is configured to:

generate a map using the second feature point and the distance for the pixel corresponding to the second feature point.

11. A motion estimation method for a motion estimation device, the method comprising:

receiving a first input frame;

estimating a distance for each pixel of the first input frame;

estimating a reliability of the distance for each pixel of the first input frame based on pixel information of the first input frame;

detecting a first feature point in the first input frame; and

selecting the first feature point as a second feature point if the estimated reliability of the distance for the pixel corresponding to first feature point is greater than or equal to a first reference value.

12. The motion estimation method according to claim 11 , further comprising:

comparing the second feature point to a third feature point in a key frame registered before the first input frame;

selecting the second feature point as a fourth feature point if the second feature point and the third feature point do not match; and

detecting the fourth feature point as a moving point in the first input frame if the fourth feature point is in a predetermined region of the first input frame.

13. The motion estimation method according to claim 12 , further comprising:

detecting the fourth feature point as a non-moving point in the first input frame if the fourth feature point is not in the predetermined region of the first input frame;

detecting the second feature point as a non-moving point in the first input frame if the second feature point and the third feature point match; and

calculating a distance to the non-moving point.

14. The motion estimation method according to claim 13 , further comprising:

comparing the estimated distance for each pixel of the first input frame to the calculated distance of the non-moving point.

15. The motion estimation method according to claim 11 , wherein the estimating of the distance for each pixel of the first input frame and the estimating of the reliability of the distance for each pixel of the first input frame is based on the pixel information of the first input frame and machine learning data.

16. The motion estimation method according to claim 11 , wherein the distance for each pixel of the first input frame is a distance between a subject and a camera that provides the first input frame.

17. The motion estimation method according claim 11 , further comprising:

receiving a second input frame, wherein

the estimating of the distance for each pixel of the first input frame and the estimating of the reliability of the estimated distance for each pixel of the first input frame is based on pixel information of the second input frame in addition to the pixel information of the first input frame.

18. A motion estimation device, comprising:

a first receiving circuit configured to receive a first input frame;

a second receiving circuit configured to receive a second input frame; and

a calculation circuit configured to perform motion estimation processing based on the first input frame, wherein

in the motion estimation processing, the calculation circuit estimates a distance for each pixel of the first input frame and estimates a reliability of the distance for each pixel of the first input frame based on pixel information of the first input frame, and

the calculation circuit, in the motion estimation processing, is further configured to:

estimate the distance for each pixel of the first input frame and estimate the reliability of the estimated distance for each pixel of the first input frame based on pixel information of the second input frame in addition to the pixel information of the first input frame.

19. The motion estimation device according to claim 18 , wherein the calculation circuit estimates the distance for each pixel of the first input frame and estimates the reliability of the distance for each pixel of the first input frame based additionally on machine learning data.

20. The motion estimation device according to claim 18 , wherein, in the motion estimation processing, the calculation circuit is configured to:

detect a first feature point in the first input frame,

select the first feature point as a second feature point if the estimated reliability of the distance for the pixel corresponding to first feature point is greater than or equal to a first reference value, and

generate a map using the second feature point and the distance for the pixel corresponding to the second feature point.