A window picture (robust window picture) is selected as a model window picture, which is unlikely to be influenced by the variations in an environment and selected irrespective of the variations in an environment of pictures assumed to be inputted in a recognition phase. Alternatively, a window picture (stable window picture) is selected as a model window picture, in which the changes in a feature value are small even though an environment is varied. For example, a tracking window picture group is obtained from a registered picture group in which an environment is continuously varied, by tracking window pictures of a basic registered picture. Then, a static extraction window picture is obtained by extracting window pictures as independent pictures from a series of pictures. Then, a window picture selected both in the tracking window picture group and in the static extraction window picture is set as a robust window picture, and picture matching is conducted using the robust window picture.

Objects are detected in an image using a spatially variant filter. The size of the filter is adjusted based on the portion of the image being processed by the filter. Portions of the image which are closer to an imaging device which captured the image are processed using a larger sized filter. Portions of the image which are further from the imaging device are processed using a smaller sized filter.

A method and apparatus for extracting feature vectors for face recognition and retrieval are provided. An entire Fourier feature vector is generated for an entire face area of a normalized face image using first and second normalized vectors, and a central Fourier feature vector is generated for a central face area of the normalized face image using third and fourth normalized vectors. An entire intensity feature vector is generated for the entire face area, and a local intensity feature vector is generated for a predetermined number of face component-wise areas. An entire composite feature vector is generated by coupling the first and second normalized vectors and the entire intensity feature, and a central composite feature vector is generated by coupling the third and fourth normalized vectors and the local intensity feature vectors.

An image input unit inputs an image including a user's face area to specify the user's gaze or face direction. A pattern generation unit generates an input pattern that can be compared with each dictionary pattern from the image. A dictionary unit previously stores a plurality of dictionary patterns each of which differently corresponds to the user's gaze or face direction. A pattern comparison unit respectively calculates a similarity degree between the input pattern and each of the dictionary patterns by comparing the input pattern with each of the dictionary patterns. A gaze detection unit detects the user's gaze or face direction based on the similarity degree.

An image processing apparatus for detecting, at high precision, the contours of the eyes and the lips from an image including a shot of the face is provided. The apparatus includes an image input portion used to input an image, a feature point detecting portion that detects plural feature points of the object from the input image, a facial pose estimating portion that estimates a rotation angle indicating a facial pose from the detected, plural feature points, an initial parameter correcting portion that estimates initial parameters to detect the contours of the eye and the lips from the detected, plural feature points, and corrects the estimated initial parameters using the estimated rotation angle, an initial contour generating portion that generates an initial contour of the object using the corrected initial parameters, and a contour generating portion that generates a final contour from the initial contour.