| 摘要 |
"OPTIMUM FAST TEXTURAL FEATURE EXTRACTOR" The invention is situated in the general field of computer image analysis; it relates to a textural parameter extractor for classification or learning. Four (4) simple 2D optimum masks independent of image, are applied to four adjacent pixels in order to diagonize the associated covariance matrice. The first three (3) central moments, namely absolute deviation, standard deviation and skewness for each transformed image constitute a feature vector of twelve (12) components for classification purposes. Parallel and Sequential structures are also presented for fast textural feature extractor applications. |
