| 摘要 |
According to the method, which can be used in real time to identify objects in an image: the image is subjected to a convolution filtering with a gaussian pulse response filter in order to obtain a transform of the image; the transform is subjected to a convolution filtering with a gaussian response filter having a wider kernel than the first filter and corresponding to an environment in which a Laplacian is significant; the grey levels which are characteristic of the interior and of the exterior of the objects are evaluated by re-updating, along a scanning line, the white or black level according to the sign of the Laplacian for this pixel; a discrimination threshold is defined by averaging, possibly weighted, between the white and black levels; and each pixel is labelled by comparison with the corresponding threshold. <IMAGE>
|
