| 摘要 |
PURPOSE: Face feature vector composition is provided to generate a face feature vector by using shape descriptors, which are generated by 2D and 3D shape models, and texture descriptors which are generated by global and local texture models. CONSTITUTION: After a face detector(110) generates an input image(105) as a landmark image(115), the landmark image is generated as 2D and 3D descriptors(210,215) through 2D and 3D shape models(200,205). A normalization image(310) which is normalized through normalization computation(305) of the landmark image is generated as a global texture descriptor(230), a local texture descriptor, and a warped or morphed image(335) through a global texture model(220), a local texture model(225), and a warping or morphing operator(330). A face feature vector(145) is generated corresponding to connection with an operator(140) and each descriptor. [Reference numerals] (110) Face detector; (145) Face feature vector; (200) 2D model; (205) 3D model; (210) 2D descriptor; (215) 3D descriptor; (220) Global texture model; (225) Local texture model; (230) Global texture descriptor; (305) Normalization; (315) Local image descriptor; (320) Dense image descriptor; (325) Warped dense image descriptor; (330) Warp; (345) Intermediate local texture descriptor; (350) Reduce dimensions |
