| 摘要 |
The starting basis for the segmentation process is the difference between two regions having different average grey values, where the edge line structure of the regions is known from previously. The grey value can vary quite widely in the image, since the processing in respect of the functional matching process which is used, constituting the smallest sum of the squares, is always based upon the image region in its entirety and not upon 5 x 3 or 5 x 5 windows, as is the case in conventional edge determination. The two-dimensional adaptation record is converted firstly into a one-dimensional minimization record, for which there are effective numerical computing algorithms, such as the process which uses the end function of the node. The need for computing work is therefore small and the process is suitable for real-time control in the process industry, including where a single- processor system is used in the image processing, provided that an edge line structure which is dependent upon one parameter is able to be adopted. <IMAGE>
|
