摘要 |
PROBLEM TO BE SOLVED: To provide a parameter estimation method capable of executing robust parameter estimation for background change and occlusion in parameter estimation of an object inside an image. SOLUTION: A feature vector of an object is extracted from digital image data where image data for learning are input (steps S101 and S102), a partial space of the object is calculated in a plurality of predetermined areas, a group of parameter estimation functions are created by support vector regression, and they are stored in a memory (steps S103 to 106). Image data of parameter estimation target are input, a feature vector is extracted from the digital image data, a group of partial spaces at the step 104 are read to project the feature vector to it, the parameter estimation function at the step S106 is read to estimate parameters in the respective areas, an optimal parameter is selected from among them, and the result is output (steps S107 to 112). COPYRIGHT: (C)2008,JPO&INPIT
|