Face detection method, device and program

A face detection method for detecting a face image contained in time-series images taken at a predetermined time interval, the method comprising the steps of:

• cutting out, for each of the time-series images, partial images having a predetermined size at different positions in the time-series image;
• calculating, for each of the partial images cut out at the different positions in each of the time-series images, an indicator value indicating a probability of the partial image being the face image;
• extracting all the partial images having the calculated indicator values not less than a first threshold value as candidates for the face image;
• setting each of the candidates as a candidate of interest, and when any nearby candidate, whose coordinate distance from the candidate of interest is not more than a predetermined distance and which is contained in a time-series image that was taken at a time within a predetermined range from the time when the time-series image containing the candidate of interest was taken, is present in the candidates other than the candidate of interest, setting the candidate of interest and the nearby candidate in one candidate group;
• calculating for each candidate group, based on the indicator values calculated for the candidates forming the candidate group, an integrated indicator value reflecting magnitudes of the indicator values;
• and
• extracting, as the face image, an image within a predetermined area containing the candidate in at least one of the time-series images that contain any of the candidates forming the candidate group having the integrated indicator value not less than a second threshold value.

A face detection device for detecting a face image contained in an inputted image, the device comprising:

• a partial image cutting means for cutting out partial images having a predetermined size at different positions in the inputted image;
• a first indicator value calculating means for calculating, for each of the partial images cut out at the different positions, an indicator value indicating a probability of the partial image being the face image;
• a face image candidate extracting means for extracting all the partial images having the calculated indicator values not less than a first threshold value as candidates for the face image;
• a candidate group setting means for setting each of the candidates as a candidate of interest, and when any nearby candidate, whose coordinate distance from the candidate of interest is not more than a predetermined distance, is present in the candidates other than the candidate of interest, setting the candidate of interest and the nearby candidate in one candidate group;
• a second indicator value calculating means for calculating for each candidate group, based on the indicator values calculated for the candidates forming the candidate group, an integrated indicator value reflecting magnitudes of the indicator values;
• and
• a face image extracting means for extracting, as the face image, an image within a predetermined area in the inputted image containing the candidate group having the integrated indicator value not less than a second threshold value image.

A face detection device for detecting a face image contained in time-series images taken at a predetermined time interval, the device comprising:

• a partial image cutting means for cutting out, for each of the time-series images, partial images having a predetermined size at different positions in the time-series image;
• a first indicator value calculating means for calculating, for each of the partial images cut out at the different positions in each of the time-series images, an indicator value indicating a probability of the partial image being the face image;
• a face image candidate extracting means for extracting all the partial images having the calculated indicator values not less than a first threshold value as candidates for the face image;
• a candidate group setting means for setting each of the candidates as a candidate of interest, and when any nearby candidate, whose coordinate distance from the candidate of interest is not more than a predetermined distance and which is contained in a time-series image that was taken at a time within a predetermined range from the time when the time-series image containing the candidate of interest was taken, is present in the candidates other than the candidate of interest, setting the candidate of interest and the nearby candidate in one candidate group;
• a second indicator value calculating means for calculating for each candidate group, based on the indicator values calculated for the candidates forming the candidate group, an integrated indicator value reflecting magnitudes of the indicator values;
• and
• a face image extracting means for extracting, as the face image, an image within a predetermined area containing the candidate in at least one of the time-series images that contain any of the candidates forming the candidate group having the integrated indicator value not less than a second threshold value.

A program for causing a computer to operate as a face detection device for detecting a face image contained in an inputted image, the program comprising causing the computer to operate as:

• a partial image cutting means for cutting out partial images having a predetermined size at different positions in the inputted image;
• a first indicator value calculating means for calculating, for each of the partial images cut out at the different positions, an indicator value indicating a probability of the partial image being the face image;
• a face image candidate extracting means for extracting all the partial images having the calculated indicator values not less than a first threshold value as candidates for the face image;
• a candidate group setting means for setting each of the candidates as a candidate of interest, and when any nearby candidate, whose coordinate distance from the candidate of interest is not more than a predetermined distance, is present in the candidates other than the candidate of interest, setting the candidate of interest and the nearby candidate in one candidate group;
• a second indicator value calculating means for calculating for each candidate group, based on the indicator values calculated for the candidates forming the candidate group, an integrated indicator value reflecting magnitudes of the indicator values;
• and
• a face image extracting means for extracting, as the face image, an image within a predetermined area in the inputted image containing the candidate group having the integrated indicator value not less than a second threshold value image.

A program for causing a computer to operate as a face detection device for detecting a face image contained in an inputted image, the program comprising causing the computer to operate as:

• a partial image cutting means for cutting out, for each of the time-series images, partial images having a predetermined size at different positions in the time-series image;
• a first indicator value calculating means for calculating, for each of the partial images cut out at the different positions in each of the time-series images, an indicator value indicating a probability of the partial image being the face image;
• a face image candidate extracting means for extracting all the partial images having the calculated indicator values not less than a first threshold value as candidates for the face image;
• a candidate group setting means for setting each of the candidates as a candidate of interest, and when any nearby candidate, whose coordinate distance from the candidate of interest is not more than a predetermined distance and which is contained in a time-series image that was taken at a time within a predetermined range from the time when the time-series image containing the candidate of interest was taken, is present in the candidates other than the candidate of interest, setting the candidate of interest and the nearby candidate in one candidate group;
• a second indicator value calculating means for calculating for each candidate group, based on the indicator values calculated for the candidates forming the candidate group, an integrated indicator value reflecting magnitudes of the indicator values;
• and
• a face image extracting means for extracting, as the face image, an image within a predetermined area containing the candidate in at least one of the time-series images that contain any of the candidates forming the candidate group having the integrated indicator value not less than a second threshold value.