| 摘要 |
PURPOSE:To easily divide a multi-gradation picture into areas by setting a target area for a sample picture, and making system learn it, and afterwards, performing the area division of the analogous multi-gradation picture. CONSTITUTION:A preprocessing circuit 2 reads out the digitized objective multi- gradation picture to be processed from a picture memory 1, and extracts the density information and the relation information of local vicinity of each picture element of this picture, and outputs these information to a calculation circuit 4. The calculation circuit 4 calculates the output of each picture element based on a calculation formula in a neural network by using these information from the preprocessing circuit 2 and a weight function from a learning circuit 3, and outputs a calculated value to a decision circuit 5. The decision circuit 5 decides whether each picture element is made to be inside the area or outside the area, and outputs a decided result to a mapping circuit 6. The mapping circuit 6 makes a displaying part 7 display the area-divided picture according to this decided result. Thus, the optimum area division can be easily executed. |
