Image recognition apparatus, image recognition method, and program
An image recognition apparatus includes an extraction unit configured to extract a feature quantity for each local area from an input image, a conversion unit configured to convert the feature quantity extracted by the extraction unit into a feature quantity indicating a degree with respect to an attribute for each local area, a verification unit configured to verify the feature quantity converted by the conversion unit against a feature quantity of a registered image, and an identification unit configured to identify whether the input image is identical to the registered image by integrating the verification result for each local area acquired by the verification unit.
1. An image recognition apparatus comprising:
an extractor that extracts a feature quantity for each local area from an input image;
a converter that converts the feature quantity for each local area extracted by the extractor into a feature quantity representing respective degrees of similarities with a plurality of representative patterns;
a verifier that verifies the feature quantity for each local area converted by the convertor against a feature quantity of a registered image, the feature quantity of the registered image representing respective degrees of similarities with the plurality of representative patterns; and
an identifier that identifies whether the input image is identical to the registered image by integrating verification results for a plurality of local areas acquired by the verification unit.
2. The image recognition apparatus according to claim 1 , wherein the converter performs the conversion by a projection using a projection matrix.
3. The image recognition apparatus according to claim 2 , wherein the projection matrix is configured such that preliminary collected image samples are classified as to whether each attribute for each local area thereof belongs to the attribute and a difference between the classified result and a result assumed based on the local area of the image sample is as small as possible.
4. The image recognition apparatus according to claim 1 , wherein the registered image is a face image of a person and the attribute is an attribute of a face part.
5. The image recognition apparatus according to claim 4 , wherein the attribute of the face part is a similarity with a face part of a representative person.
6. An image recognition method comprising:
extracting a feature quantity for each local area from an input image;
converting the extracted feature quantity for each local area into a feature quantity representing respective degrees of similarities with a plurality of representative patterns;
verifying the converted feature quantity for each local area against a feature quantity of a registered image, the feature quantity of the registered image representing respective degrees of similarities with the representative patterns; and
identifying whether the input image is identical to the registered image by integrating verification results for each local area.
7. A non-transitory computer-readable storage medium storing a control program for causing a computer to execute each operation of the image recognition method according to claim 6 .