Three-dimensional face model generation method and apparatus, device, and medium

1. A three-dimensional face model generation method, applied to a computing device, the method comprising:

obtaining a three-dimensional face mesh of a target object;

aligning the three-dimensional face mesh with a first three-dimensional face model of a standard object according to face keypoints;

performing local fitting on the three-dimensional face mesh and a local area of the first three-dimensional face model, to obtain a second three-dimensional face model after the local fitting, wherein the local area is an area whose fitting error rate is lower than a preset condition, the three-dimensional face mesh further includes a candidate area whose fitting error rate is higher than or equal to the preset condition, and the local fitting is not performed on the candidate area; and

performing global fitting on the three-dimensional face mesh and a global area of the second three-dimensional face model, to obtain a three-dimensional face model of the target object after the global fitting, wherein the global fitting is performed on the candidate area to obtain the 3D face model of the target object.

2. The method according to claim 1 , wherein performing local fitting on the three-dimensional face mesh and the local area of the first three-dimensional face model comprises:

performing registration on the three-dimensional face mesh and the local area of the first three-dimensional face model, to obtain a first correspondence between the local area and the three-dimensional face mesh; and

performing the local fitting on the three-dimensional face mesh and the local area of the first three-dimensional face model according to the first correspondence, to obtain the second three-dimensional face model after the local fitting.

3. The method according to claim 2 , wherein performing local fitting on the three-dimensional face mesh and the local area of the first three-dimensional face model comprises:

performing the local fitting on the three-dimensional face mesh and the local area of the first three-dimensional face model according to the first correspondence, to calculate a first posture parameter of the second three-dimensional face model;

adjusting a shape base coefficient of the first three-dimensional face model according to the first posture parameter, to obtain a shape base coefficient of the second three-dimensional face model;

adjusting an expression base coefficient of the first three-dimensional face model according to the first posture parameter, to obtain an expression base coefficient of the second three-dimensional face model; and

obtaining, according to the first posture parameter, the shape base coefficient, and the expression base coefficient, the second three-dimensional face model after the local fitting.

4. The method according to claim 1 , wherein performing global fitting on the three-dimensional face mesh and the global area of the second three-dimensional face model comprises:

performing registration on the three-dimensional face mesh and the global area of the second three-dimensional face model, to obtain a second correspondence between the global area and the three-dimensional face mesh; and

performing the global fitting on the three-dimensional face mesh and the global area of the second three-dimensional face model according to the second correspondence, to obtain the three-dimensional face model of the target object after the global fitting.

5. The method according to claim 4 , wherein performing global fitting on the three-dimensional face mesh and the global area of the second three-dimensional face model comprises:

performing the global fitting on the three-dimensional face mesh and the global area of the second three-dimensional face model according to the second correspondence, to calculate a second posture parameter of the three-dimensional face model of the target object after the global fitting;

adjusting a shape base coefficient of the second three-dimensional face model according to the second posture parameter, to obtain a shape base coefficient of the three-dimensional face model of the target object after the global fitting;

adjusting an expression base coefficient of the second three-dimensional face model according to the second posture parameter, to obtain an expression base coefficient of the three-dimensional face model of the target object after the global fitting; and

obtaining, according to the second posture parameter, the shape base coefficient, and the expression base coefficient, the three-dimensional face model of the target object after the global fitting.

6. The method according to claim 4 , further comprising:

obtaining a flaw of the second three-dimensional face model, wherein the flaw corresponds to an outlier of the first three-dimensional face model, and the outlier is a point that does not belong to the local area of the first three-dimensional face model; and

filtering out the flaw from the second three-dimensional face model, to obtain a second three-dimensional face model that is flawless.

7. The method according to claim 6 , wherein obtaining the flaw of the second three-dimensional face model comprises:

comparing a norm corresponding to the global area of the second three-dimensional face model with a norm corresponding to the second three-dimensional face model after the local fitting, to obtain a norm difference; and

determining the flaw of the second three-dimensional face model according to the norm difference.

8. The method according to claim 1 , further comprising:

obtaining a shape base coefficient and an expression base coefficient of the three-dimensional face model of the target object after the global fitting; and

generating a three-dimensional expression base of the target object according to the shape base coefficient and the expression base coefficient.

9. A three-dimensional face model generation apparatus, the apparatus comprising: at least one memory storing computer program instructions; and at least one processor coupled to the at least one memory and configured to execute the computer program instructions and perform:

obtaining a three-dimensional face mesh of a target object;

aligning the three-dimensional face mesh with a first three-dimensional face model of a standard object according to face keypoints; and

performing local fitting on the three-dimensional face mesh and a local area of the first three-dimensional face model, to obtain a second three-dimensional face model after the local fitting, wherein the local area is an area whose fitting error rate is lower than a preset condition, the three-dimensional face mesh further includes a candidate area whose fitting error rate is higher than or equal to the preset condition, and the local fitting is not performed on the candidate area; and

performing global fitting on the three-dimensional face mesh and a global area of the second three-dimensional face model, to obtain a three-dimensional face model of the target object after the global fitting, wherein the global fitting is performed on the candidate area to obtain the 3D face model of the target object.

10. The apparatus according to claim 9 , wherein performing local fitting on the three-dimensional face mesh and the local area of the first three-dimensional face model includes:

performing registration on the three-dimensional face mesh and the local area of the first three-dimensional face model, to obtain a first correspondence between the local area and the three-dimensional face mesh; and

performing the local fitting on the three-dimensional face mesh and the local area of the first three-dimensional face model according to the first correspondence, to obtain the second three-dimensional face model after the local fitting.

11. The apparatus according to claim 10 , wherein performing local fitting on the three-dimensional face mesh and the local area of the first three-dimensional face model includes:

performing the local fitting on the three-dimensional face mesh and the local area of the first three-dimensional face model according to the first correspondence, to calculate a first posture parameter of the second three-dimensional face model;

adjusting a shape base coefficient of the first three-dimensional face model according to the first posture parameter, to obtain a shape base coefficient of the second three-dimensional face model;

adjusting an expression base coefficient of the first three-dimensional face model according to the first posture parameter, to obtain an expression base coefficient of the second three-dimensional face model; and

obtaining, according to the first posture parameter, the shape base coefficient, and the expression base coefficient, the second three-dimensional face model after the local fitting.

12. The apparatus according to claim 9 , wherein performing global fitting on the three-dimensional face mesh and the global area of the second three-dimensional face model includes:

performing registration on the three-dimensional face mesh and the global area of the second three-dimensional face model, to obtain a second correspondence between the global area and the three-dimensional face mesh; and

performing the global fitting on the three-dimensional face mesh and the global area of the second three-dimensional face model according to the second correspondence, to obtain the three-dimensional face model of the target object after the global fitting.

13. The apparatus according to claim 12 , wherein performing global fitting on the three-dimensional face mesh and the global area of the second three-dimensional face model includes:

performing the global fitting on the three-dimensional face mesh and the global area of the second three-dimensional face model according to the second correspondence, to calculate a second posture parameter of the three-dimensional face model of the target object after the global fitting;

adjusting a shape base coefficient of the second three-dimensional face model according to the second posture parameter, to obtain a shape base coefficient of the three-dimensional face model of the target object after the global fitting;

adjusting an expression base coefficient of the second three-dimensional face model according to the second posture parameter, to obtain an expression base coefficient of the three-dimensional face model of the target object after the global fitting; and

obtaining, according to the second posture parameter, the shape base coefficient, and the expression base coefficient, the three-dimensional face model of the target object after the global fitting.

14. The apparatus according to claim 12 , wherein the at least one processor is further configured to perform:

obtaining a flaw of the second three-dimensional face model, wherein the flaw corresponds to an outlier of the first three-dimensional face model, and the outlier is a point that does not belong to the local area of the first three-dimensional face model; and

filtering out the flaw from the second three-dimensional face model, to obtain a second three-dimensional face model that is flawless.

15. The apparatus according to claim 14 , wherein obtaining the flaw of the second three-dimensional face model includes:

comparing a norm corresponding to the global area of the second three-dimensional face model with a norm corresponding to the second three-dimensional face model after the local fitting, to obtain a norm difference; and

determining the flaw of the second three-dimensional face model according to the norm difference.

16. The apparatus according to claim 9 , wherein the at least one processor is further configured to perform:

obtaining a shape base coefficient and an expression base coefficient of the three-dimensional face model of the target object after the global fitting; and

generating a three-dimensional expression base of the target object according to the shape base coefficient and the expression base coefficient.

17. A non-transitory computer-readable storage medium storing computer program instructions executable by at least one processor to perform:

obtaining a three-dimensional face mesh of a target object;

aligning the three-dimensional face mesh with a first three-dimensional face model of a standard object according to face keypoints;

performing local fitting on the three-dimensional face mesh and a local area of the first three-dimensional face model, to obtain a second three-dimensional face model after the local fitting, wherein the local area is an area whose fitting error rate is lower than a preset condition, the three-dimensional face mesh further includes a candidate area whose fitting error rate is higher than or equal to the preset condition, and the local fitting is not performed on the candidate area; and

performing global fitting on the three-dimensional face mesh and a global area of the second three-dimensional face model, to obtain a three-dimensional face model of the target object after the global fitting, wherein the global fitting is performed on the candidate area to obtain the 3D face model of the target object.

18. The non-transitory computer-readable storage medium according to claim 17 , wherein performing local fitting on the three-dimensional face mesh and the local area of the first three-dimensional face model includes:

performing registration on the three-dimensional face mesh and the local area of the first three-dimensional face model, to obtain a first correspondence between the local area and the three-dimensional face mesh; and

performing the local fitting on the three-dimensional face mesh and the local area of the first three-dimensional face model according to the first correspondence, to obtain the second three-dimensional face model after the local fitting.

19. The method according to claim 1 , wherein a quantity of face keypoints of the second three-dimensional face model is higher than a quantity of face keypoints of the first three-dimensional face model.

20. The method according to claim 1 , wherein the candidate area includes at least one of a nose area or a mouth area.