| 摘要 |
<P>PROBLEM TO BE SOLVED: To suppress a discrimination accuracy reduction due to the effects of superposition on a face on an image, oblique lighting, density variations in the background other than the face, and the like, in the process of normalizing the contrast of the image and then discriminating whether or not the image is a face image according to a feature of brightness distribution on the image. <P>SOLUTION: The normalization process is the process of pixel value gradation conversion that approximates the degree of variance of pixel values to a predetermined level in each local region of a discrimination target image, and the local region is a region sized to include only one eye of a face to be discriminated. The normalization can be stabilized in which contrast changes are reduced in face components featuring the face, such as the eyes and nose, and the effect of density changes by components other than the face is reduced. <P>COPYRIGHT: (C)2007,JPO&INPIT |
