| 摘要 |
PURPOSE: To improve noise control by deciding the motion vector of a detected object by minimizing a mean-square Gray value difference over the intersection set of deviated mask segments and utilizing the difference between the centroid vectors of corresponding segments as a starting vector at the time of minimizing. CONSTITUTION: One column of binary object masks are calculated by the binarization of a difference image column from an input image stream and a background image column through the use of one threshold value, those threshold values are used for controlling spatial selectivity at the time of calculating the background image column and those segments are decided with their dimension and centered position. Then the motion vector of the detected object is decided by minimizing the means-square Gray value difference over the intersection set of the deviated mask segments and the difference between the centroid vectors of the corresponding segment is utilized as the starting vector at the time. Thereby necessary noise suppression in the object mask is executed by the application of a median filter. |
