| 摘要 |
PROBLEM TO BE SOLVED: To detect, recognize, and encode various objects like human face. SOLUTION: This device is provided with a means for detecting an object with target characteristics by scanning an image with various contraction scales and preferably, by performing maximum likli-hood analysis in parallel, a means, preferably, expressing a directional image set with coincidence orientation and/or illuminance as a characteristic vector set in multi-dimensional space, a means for adjusting the detected component of an inputted image so as to obtain coincidence directional image set, geometry, and/or gray scale contrast, a means for projecting the detected characteristics on an intrinsic space expression of the directional image set for recognition and arbitrary reconfiguration module 210 for reconfiguring an original image in relation to the characteristic vector projection coefficient, and an adjusted and/or the reconfigured image is efficiently expressed as a set of a quantized characteristic vector projection coefficient. |
