| 摘要 |
<P>PROBLEM TO BE SOLVED: To automatically correct a teaching model to be used in detecting an object from an input image. Ž<P>SOLUTION: The image processor 10 detects an image of a detection object from the input image based on correlation between the featured value model of the object of detection preliminarily defined by using a plurality of featured values and an input image. A featured value contribution exponential arithmetic means performs correlation arithmetic processing to calculate correlation, and calculates the degree of contribution to the increase of correlation in previous correlation arithmetic processing about each featured value of a featured value model, and searches a contribution exponential based on a value calculated by accumulating the calculated degree of contribution. A featured value model correction means 40 corrects the featured value model by determining whether or not the contribution exponential calculated by the featured value contribution exponential calculation means is equal to or less than a predetermined threshold about each featured value of the featured value model, and removing the featured value from the a plurality of the featured values of the featured value model when the contribution exponential is equal to or less than the predetermined threshold. Ž<P>COPYRIGHT: (C)2010,JPO&INPIT Ž
|
