| 摘要 |
PURPOSE:To easily restore the contour line of the pattern and to obtain a restored pattern of high quality by utilizing the learning function of a neural network(NN) for the approximate expression of the contour line of the pattern and utilizing the interpolating function of the NN for the restoration and generation of the contour line. CONSTITUTION:An approximation section setting means 1 sets an approximation section wherein the contour line of the original pattern is approximated with a straight line or curved line. A straight line/curved line approximating means 2 includes the NN, learns two-dimensional coordinate values on the contour line consisting of the straight line or curved line in the respective approximation sections set by the means 1 as learning data on the NN, and stores the relation between the two-dimensional coordinate values of all contour points on the contour line as the weight of the coupling of the NN. A contour point sequence generating means 4 restores the contour line and performs size conversion as to the respective contour points in the restored contour point sequence to restore the contour line. A pattern restoring and generating means 5 paints the inside of the restored contour out to obtain the same restored or generated pattern as the original pattern. |
