Information processing device for generating a face image with converting attribute of the fame image, image processing method for implementing the same, and recording medium for implementing the same

1 . An image processing device comprising:

a memory configured to store instructions; and

one or more processors configured to execute the instructions to:

calculate a first similarity score group and m second similarity score groups, the first similarity score group including n similarity scores indicating similarity between a first face image corresponding to a face image of a target person and each of n (n≥2) face images, the m second similarity score groups including n similarity scores indicating similarity between m second face images and each of the n face images, the m second face images being obtained by converting an attribute of the first face image by individually applying m (m≥2) conversion methods to the first face image;

acquire a first score distribution and m second score distributions, the first score distribution indicating a distribution state of at least a part of the similarity scores in the first similarity score group, the m second score distributions indicating a distribution state of at least a part of the similarity scores in each of the m second similarity score groups;

acquire at least one conversion parameter corresponding to at least one conversion method from among the m conversion methods based on comparison results between the first score distribution and the m second score distributions, and

generate a face image by converting the attribute of the first face image based on the conversion parameter.

2 . The image processing device according to claim 1 ,

wherein the processor is further configured to execute the instructions to calculate m distances between the first score distribution and the m second score distributions, and

wherein the processor acquires the conversion parameter by comparing calculation results of the m distances.

3 . The image processing device according to claim 2 , wherein the processor extracts one score distribution for which the distance from the first score distribution becomes minimum from the m second score distributions, based on the calculation results of the m distances, and acquires one conversion parameter corresponding to one conversion method applied to the first face image for acquiring the one score distribution.

4 . The image processing device according to claim 2 , wherein the processor extracts a plurality of score distributions from the m second score distributions in order of smaller distance from the first score distribution based on the calculation results of the m distances, and acquires a plurality of conversion parameters corresponding to the plurality of conversion methods applied to the first face image to acquire the plurality of score distributions.

5 . The image processing device according to claim 1 , wherein the processor uses p (2≤p≤n) similarity scores calculated from common p face images among the n face images to acquire the first score distribution and the m second score distributions.

6 . The image processing device according to claim 5 , wherein the processor acquires the first score distribution indicating the distribution state of top k (2≤k≤p) similarity scores among the p similarity scores, and acquires the m second score distributions indicating the distribution state of the top k similarity scores among the p similarity scores.

7 . An image processing method comprising:

calculating a first similarity score group and m second similarity score groups, the first similarity score group including n similarity scores indicating similarity between a first face image corresponding to a face image of a target person and each of n (n≥2) face images, the m second similarity score groups including n similarity scores indicating similarity between m second face images and each of the n face images, the m second face images being obtained by converting an attribute of the first face image by individually applying m (m≥2) conversion methods to the first face image;

acquiring a first score distribution and m second score distributions, the first score distribution indicating a distribution state of at least a part of the similarity scores in the first similarity score group, the m second score distributions indicating a distribution state of at least a part of the similarity scores in each of the m second similarity score groups;

acquiring at least one conversion parameter corresponding to at least one conversion method from among the m conversion methods based on comparison results between the first score distribution and the m second score distributions, and

generating a face image by converting the attribute of the first face image based on the conversion parameter.

8 . A non-transitory computer-readable recording medium recording a program, the program causing a computer to execute processing comprising:

calculating a first similarity score group and m second similarity score groups, the first similarity score group including n similarity scores indicating similarity between a first face image corresponding to a face image of a target person and each of n (n≥2) face images, the m second similarity score groups including n similarity scores indicating similarity between m second face images and each of the n face images, the m second face images being obtained by converting an attribute of the first face image by individually applying m (m≥2) conversion methods to the first face image;

acquiring a first score distribution and m second score distributions, the first score distribution indicating a distribution state of at least a part of the similarity scores in the first similarity score group, the m second score distributions indicating a distribution state of at least a part of the similarity scores in each of the m second similarity score groups;

acquiring at least one conversion parameter corresponding to at least one conversion method from among the m conversion methods based on comparison results between the first score distribution and the m second score distributions, and

generating a face image by converting the attribute of the first face image based on the conversion parameter.