Measurement-target-selecting device, face-shape-estimating device, method for selecting measurement target, and method for estimating face shape

1. A measurement target extracting apparatus that extracts a candidate for a three-dimensional shape measurement target among a plurality of face shape candidates, the apparatus comprising:

a microcomputer arranged to:

calculate, for each of the plurality of face shape candidates, a facial texture evaluation value representing a degree of match between a texture of an input face image and a texture of the face shape candidate;

calculate, for each of the plurality of face shape candidates, first likelihood between a reference face shape and the face shape candidate as a facial expression changing likelihood;

calculate, for each of the plurality of face shape candidates, a first correlation evaluation value representing a degree of correlation between the facial texture evaluation value and the first likelihood; and

extract, from the plurality of face shape candidates, a face shape candidate having the first correlation evaluation value lower than a first threshold, as the measurement target.

2. The measurement target extracting apparatus according to claim 1 , wherein the microcomputer is further arranged to:

calculate the facial texture evaluation value, and calculates, for each of face parts of the face shape candidate, a face part texture evaluation value representing a degree of match between the facial image and the face part;

calculate the first likelihood, and calculate, for each of the face parts in the face shape candidate, second likelihood between the reference face shape and the face part;

calculate the first correlation evaluation value, and calculate, for each of the face parts in the face shape candidate, a second correlation evaluation value representing a degree of correlation between the face part texture evaluation value and the second likelihood; and

extract, from face parts in a face shape candidate having the first correlation evaluation value lower than the first threshold, a face part having a second correlation value lower than a second threshold, as a measurement target.

3. The measurement target extracting apparatus according to claim 2 , wherein the microcomputer is further arranged to select, from face parts having the second correlation evaluation values lower than the second threshold, a number of face parts corresponding to a predetermined proportion of to a total number of the face parts in ascending order of the second correlation values, and extract the selected face parts as the measurement target.

4. The measurement target extracting apparatus according to claim 2 , wherein the microcomputer is further arranged to calculate the first likelihood representing a degree of match between the position of the face part in the face shape candidate and a movable range of mimic muscles for moving the face part, by comparing the position of the face part in the face shape candidate and a position of the reference face shape.

5. The measurement target extracting apparatus according to claim 1 , wherein the microcomputer is further arranged to calculate, for each of the plurality of face shape candidates, the first likelihood representing an amount of change between three-dimensional coordinates of a plurality of feature points representing the face shape candidate and three-dimensional coordinates at positions corresponding to the plurality of feature points in the reference face shape.

6. The measurement target extracting apparatus according to claim 1 , wherein the reference face shape is an expressionless face shape.

7. The measurement target extracting apparatus according to claim 1 , wherein the reference face shape is a face shape obtained a specific time ago.

8. A face shape estimating apparatus comprising:

the measurement target extracting apparatus according to claim 1 ; and

the microcomputer or another microcomputer being arranged to,

measure a three-dimensional shape of a face shape candidate extracted as the measurement target by the measurement target extracting apparatus;

calculate a first shape candidate evaluation value for each of the plurality of face shape candidates, based on the facial texture evaluation value and the first likelihood;

calculate, for each of the plurality of extracted face shape candidates, a second shape candidate evaluation value, using the three-dimensional shape; and

determine an estimated shape of the facial image from the plurality of face shape candidates based on the first shape candidate evaluation value and the second shape candidate evaluation value.

9. The face shape estimating apparatus according to claim 8 , wherein:

in the measurement target extracting apparatus, the microcomputer is further arranged to:

calculate the face texture evaluation value, and calculate, for each of face parts in the face shape candidate, a face part texture evaluation value representing a degree of match between the facial image and the face part;

calculate the first likelihood, and calculate, for each of the face parts in the face shape candidate, second likelihood between the reference face shape and the face part;

calculate the first correlation evaluation value, and calculate, for each of the face parts in the face shape candidate, a second correlation evaluation value representing a degree of correlation between the face part texture evaluation value and the second likelihood; and

extract, from face parts in a face shape candidate having the first correlation evaluation value lower than the first threshold, a face part having the second correlation value lower than a second threshold, as the measurement target,

the microcomputer or the another microcomputer being further arranged to:

measure a three-dimensional shape of the face part that was extracted; and

calculate the second shape candidate evaluation value, using the three-dimensional shape of the face part.

10. The face shape estimating apparatus according to claim 8 , wherein the microcomputer or the another microcomputer is further arranged to calculate, for each of the extracted face shape candidates, the second shape candidate evaluation value based on the facial texture evaluation value and a result of comparison between the three-dimensional shape and three-dimensional coordinates of the face shape candidate.

11. The face shape estimating apparatus according to claim 8 , wherein the microcomputer or the another microcomputer is further arranged to generate a new face shape candidate using the extracted face shape candidate and information regarding the three-dimensional shape, and calculate the second shape candidate evaluation value corresponding to each of the new face shape candidates, based on the facial texture evaluation value and the first likelihood that correspond to the new face shape candidate.

12. The face shape estimating apparatus according to claim 8 , wherein the microcomputer or the another microcomputer is further arranged to determine, as the estimated shape, a face shape candidate corresponding to the largest value among the first shape candidate evaluation value and the second shape candidate evaluation value.

13. The face shape estimating apparatus according to claim 8 , wherein the microcomputer or the another microcomputer is further arranged to determine the estimated shape by calculating a weighted average of the plurality of face shape candidates, based on a weight calculated using the first shape candidate evaluation value and the second shape candidate evaluation value.

14. A measurement target extracting method for extracting a candidate for a three-dimensional shape measurement target from a plurality of face shape candidates, the method comprising:

calculating, for each of the plurality of face shape candidates, a facial texture evaluation value representing a degree of match between a texture of an input face image and a texture of the face shape candidate;

calculating, for each of the plurality of face shape candidates, first likelihood between a reference face shape and the face shape candidate;

calculating, for each of the plurality of face shape candidates, a first correlation evaluation value representing a degree of correlation between the facial texture evaluation value and the first likelihood; and

extracting, from the plurality of face shape candidates, a face shape candidate having the first correlation evaluation value lower than a first threshold, as the measurement target.

15. A face shape estimating method comprising:

measuring a three-dimensional shape of a face shape candidate extracted as the measurement target by the measurement target extracting method according to claim 14 ;

calculating a first shape candidate evaluation value for each of the plurality of face shape candidates, based on the facial texture evaluation value and the first likelihood;

calculating, for each of the plurality of extracted face shape candidates, a second shape candidate evaluation value, using the three-dimensional shape; and

determining an estimated shape of the facial image from the plurality of face shape candidates based on the first shape candidate evaluation value and the second shape candidate evaluation value.