| 摘要 |
A method for fast locating a decipherable pattern in an input image, which is characterized in utilizing an overly downscaled binary image to not only reduce computation time but also facilitate extraction of skeletons for fast and accurately locating pattern, is disclosed. First, a pre-process is applied to an input image to acquire a binary image downscaled n times, from which at least one skeleton corresponding to a decipherable pattern is extracted. Coordinate values of at least one pixel of each skeleton are respectively enlarged n1/2 times and used as the central points on the original image plane for establishing a plurality of detecting blocks with the identical size. Subsequently, a grading mechanism is employed to determine the corresponding detecting blocks of the decipherable pattern.
|
