| 摘要 |
<P>PROBLEM TO BE SOLVED: To provide an image processing device capable of accurately identifying a plurality of areas within an image. <P>SOLUTION: An eye recognizing part 101 recognizes the positions of both eyes in a face image stored in an image memory 201. A skin/hair sample area determining part 102 determines a skin sample area and a hair sample area. A skin/hair sample acquisition part 103 acquires a skin sample and a hair sample. A skin composite model generating part 104 calculates the average vector and the covariance matrix of a skin area and a composite area resulting from the combining of the skin area and a hair area and creates a skin model and a composite model. A probability density function computing part 105 calculates the probability that pixel data belongs to the skin area and the probability that the data belongs to the composite area. An area identifying part 106 identifies the skin area, the composite area and a background area. A hair area identifying part 107 obtains the hair area by subtracting the skin area from the composite area. A skin composite model updating part 109 newly calculates the average vector and the covariance matrix to update the skin/hair models. <P>COPYRIGHT: (C)2005,JPO&NCIPI |
