Image processing apparatus and method, and storage medium

1. An image-processing apparatus comprising:

a face region extraction unit configured to extract face regions of a person from input images;

a calculation unit configured to calculate facial expression evaluation values from the extracted face regions using an evaluation formula;

a determination unit configured to determine facial expressions represented by the face regions based on the calculated facial expression evaluation values;

a measurement unit configured to measure a time series change in the calculated facial expression evaluation values corresponding to a plurality of the images input during a predetermined time period; and

a change unit configured to change, in a case that a difference between a maximum value and a minimum value of the calculated facial expression evaluation values calculated during the predetermined time period is smaller than a predetermined value, one of the evaluation formula and a parameter of the evaluation formula to increase the difference.

2. The apparatus according to claim 1 , wherein said change unit changes one of the evaluation formula and the parameter of the evaluation formula to make the maximum value and the minimum value of the calculated facial expression evaluation values take predetermined values.

3. The apparatus according to claim 1 , further comprising a partial region extraction unit configured to extract, from the extracted face regions, a predetermined partial region in which the facial expression changes between a first state and a second state,

wherein said calculation unit calculates the facial expression evaluation values using, for the extracted partial region, an evaluation formula which yields a first value in a case that the facial expression indicates a likelihood of the occurrence of the second state and a second value, smaller than the first value, in a case that the facial expression indicates a likelihood of the occurrence of the first state, and

wherein said determination unit determines that the facial expression indicates the second state in a case that the calculated facial expression evaluation value exceeds a threshold value.

4. The apparatus according to claim 3 , wherein said calculation unit has a neural network whose neuron output value becomes larger in the second state than in the first state, and the evaluation formula represents a relationship between the facial expression evaluation value and the number of neuron output values not less than a threshold value in the neuron output values of pixels of the partial region.

5. The apparatus according to claim 4 , wherein the relationship between the facial expression evaluation value and the number of neuron output values not less than the threshold value in the neuron output values of the pixels of the partial region is represented by a hyperbolic tangent function.

6. The apparatus according to claim 4 , wherein said change unit changes the threshold value as the parameter.

7. The apparatus according to claim 3 , wherein

the predetermined partial region is a pupil region,

the first state is an eye open state, and

the second state is an eye closed state.

8. An image-processing method comprising:

extracting face regions of a person from input images;

calculating facial expression evaluation values from the extracted face regions using an evaluation formula;

determining facial expressions represented by the face regions based on the calculated facial expression evaluation values;

measuring a time series change in the calculated facial expression evaluation values corresponding to a plurality of the images input during a predetermined time period; and

in a case that a difference between a maximum value and a minimum value of the calculated facial expression evaluation values calculated during the predetermined time period is smaller than a predetermined value, changing one of the evaluation formula and a parameter of the evaluation formula to increase the difference.

9. A non-transitory computer-readable storage medium which stores a program for causing an image-processing apparatus to:

extract face regions of a person from input images;

calculate facial expression evaluation values from the extracted face regions using an evaluation formula;

determine facial expressions represented by the face regions based on the calculated facial expression evaluation values;

measure a time series change in the calculated facial expression evaluation values corresponding to a plurality of the images input during a predetermined time period; and

in a case that a difference between a maximum value and a minimum value of the calculated facial expression evaluation values calculated during the predetermined time period is smaller than a predetermined value, change one of the evaluation formula and a parameter of the evaluation formula to increase the difference.

10. An image-processing apparatus comprising:

face region extraction means for extracting face regions of a person from input images;

calculation means for calculating facial expression evaluation values from the extracted face regions using an evaluation formula;

determination means for determining facial expressions represented by the face regions based on the calculated facial expression evaluation values;

measuring means for measuring a time series change in the calculated facial expression evaluation values corresponding to a plurality of images input during a predetermined time period; and

change means for, in a case that a difference between a maximum value and a minimum value of the calculated facial expression evaluation values calculated during the predetermined time period is smaller than a predetermined value, changing one of the evaluation formula and a parameter of the evaluation formula to increase the difference.