3D human face reconstruction method, apparatus and server

1. A three-dimensional (3D) human face reconstruction method, comprising:

obtaining a two-dimensional (2D) human face image for 3D human face reconstruction, and determining feature points on the 2D human face image, wherein the feature points are representative of a human face contour;

determining posture parameters of a human face by using the feature points, and adjusting, based on the posture parameters, a posture of a general 3D human face model obtained in advance;

determining points on the general 3D human face model corresponding to the feature points, and adjusting corresponding points in a blocking state to obtain a preliminary 3D human face model;

performing deformation adjusting on the preliminary 3D human face model to obtain a deformed 3D human face model, so that a positional relationship among the corresponding points on the deformed 3D human face model is consistent with a positional relationship among the feature points on the 2D human face image; and

performing texture mapping on the deformed 3D human face model to obtain a 3D human face,

wherein adjusting the corresponding points in the blocking state to obtain the preliminary 3D human face model comprises:

determining a plane in which the corresponding points in the blocking state are located, the plane being perpendicular to the longitudinal axis of the human face;

determining an intersection track between the plane and the general 3D human face model; and

moving the corresponding points to the outermost end of the intersection track, and

wherein the method is performed by one or more processors.

2. The method according to claim 1 , wherein performing deformation adjusting on the preliminary 3D human face model comprises:

calculating displacements of the corresponding points on the preliminary 3D human face model with respect to the feature points, with reference to a proportional relation among the feature points on the 2D human face image;

constructing an interpolation function, and calculating displacements of other points on the preliminary 3D human face model except the corresponding points based on the interpolation function; and

adjusting the preliminary 3D human face model according to the displacements of the corresponding points and the displacements of the non-corresponding points.

3. The method according to claim 2 , wherein constructing the interpolation function comprises constructing the interpolation function by using the radial basis function.

4. The method according to claim 1 , wherein performing texture mapping on the deformed 3D human face model comprises performing the texture mapping by using a similarity-preserving mesh parameterization method.

5. A three-dimensional (3D) human face reconstruction apparatus, comprising:

one or more processors configured to:

obtain a two-dimensional (2D) human face image for 3D human face reconstruction, and determine feature points on the 2D human face image, wherein the feature points are representative of a human face contour;

determine posture parameters of a human face by using the feature points, and adjust, based on the posture parameters, a posture of a general 3D human face model obtained in advance;

determine points on the general 3D human face model corresponding to the feature points, and adjust corresponding points in a blocking state to obtain a preliminary 3D human face model;

perform deformation adjusting on the preliminary 3D human face model to obtain a deformed 3D human face model, so that a positional relationship among the corresponding points on the deformed 3D human face model is consistent with a positional relationship among the feature points on the 2D human face image; and

perform texture mapping on the deformed 3D human face model to obtain a 3D human face,

wherein the one or more processors are further configured to:

determine a plane in which the corresponding points in the blocking state are located, the plane being perpendicular to the longitudinal axis of the human face;

determine an intersection track between the plane and the general 3D human face model; and

move the corresponding points to the outermost end of the intersection track.

6. The 3D human face reconstruction apparatus according to claim 5 , wherein the one or more processors are further configured to:

calculate displacements of the corresponding points on the preliminary 3D human face model with respect to the feature points, with reference to a proportional relation among the feature points on the 2D human face image;

construct an interpolation function, and calculate displacements of other points on the preliminary 3D human face model except the corresponding points based on the interpolation function; and

adjust the preliminary 3D human face model according to the displacements of the corresponding points and the displacements of the non-corresponding points.

7. The 3D human face reconstruction apparatus according to claim 6 , wherein the interpolation function is constructed by using the radial basis function.

8. The 3D human face reconstruction apparatus according to claim 5 , wherein the texture mapping is performed on the deformed 3D human face model based on similarity-preserving mesh parameterization.

9. A server, comprising a three-dimensional (3D) human face reconstruction apparatus, the 3D human face reconstruction apparatus comprising one or more processors and a memory for storing program instructions that, when executed by the one or more processors, cause the one or more processors to:

obtain a two-dimensional (2D) human face image for 3D human face reconstruction, and determine feature points on the 2D human face image, wherein the feature points are representative of a human face contour;

determine posture parameters of a human face by using the feature points, and adjust, based on the posture parameters, a posture of a general 3D human face model obtained in advance;

determine points on the general 3D human face model corresponding to the feature points, and adjust corresponding points in a blocking state to obtain a preliminary 3D human face model;

perform deformation adjusting on the preliminary 3D human face model to obtain a deformed 3D human face model, so that a positional relationship among the corresponding points on the deformed 3D human face model is consistent with a positional relationship among the feature points on the 2D human face image; and

perform texture mapping on the deformed 3D human face model to obtain a 3D human face,

wherein the program instructions further cause the one or more processors to:

determine a plane in which the corresponding points in the blocking state are located, with the plane being perpendicular to the longitudinal axis of the human face;

determine an intersection track between the plane and the general 3D human face model; and

move the corresponding points to the outermost end of the intersection track.

10. The server according to claim 9 , wherein the program instructions further cause the one or more processors to:

calculate displacements of the corresponding points on the preliminary 3D human face model with respect to the feature points, with reference to a proportional relation among the feature points on the 2D human face image;

construct an interpolation function, and calculate displacements of other points on the preliminary 3D human face model except the corresponding points based on the interpolation function; and

adjust the preliminary 3D human face model according to the displacements of the corresponding points and the displacements of the non-corresponding points.

11. The server according to claim 10 , wherein the interpolation function is constructed by using the radial basis function.

12. The server according to claim 9 , wherein the texture mapping is performed on the deformed 3D human face model based on similarity-preserving mesh parameterization.

13. A non-transitory computer readable storage medium, comprising program instructions which, when executed by a processor of a computing device, configure the processor to execute the 3D human face reconstruction method according to claim 1 .