| 摘要 |
Each image is swept, sampled and digitised, line by line and point by point, each image point being represented by the value of its luminance. During an image sweep, points, called movement points, whose luminance value is different to that of their equivalent points in a preceding image, called the reference image, with a difference greater than a fixed step, is attributed to a zone each zone regrouping connected points in movement is given a number (Zi). Each zone (Zi) is characterised by the components of a vector E(Zi), called the state vector, which contains the number N(Zi) of points attributed to the zone (Zi). For each zone of the current image (Zi) a search is made to determine if there exists equivalent zones in the zone (Zi) of the preceding current image, from which is deduced a value, called the former value of this zone. It is considered that a point is in true movement, not due to noise, if the zone to which it belongs, contains a number of points greater than a fixed step and if it has a former value greater than a fixed step. |
