| 摘要 |
<p>PROBLEM TO BE SOLVED: To provide an processing method which obtains an output image having smooth edge parts and oblique lines for any input image at the time of converting the resolution and magnification of the image. SOLUTION: A image data input meats 1 which inputs an n-gradation (n>1) original image, a image data conversion means 2 which subjects the inputted original image to resolution or magnification conversion processing, a data output means 3 which outputs a converted are provided. The image data conversion means 2 uses a-hierarchical neural network capable of learning for each input image, to perform the conversion processing of each n-gradation input image data divided into plural matrices. Thus, the most suitable conversion processing is always performed, because the weight of the network can be adjusted for each input image, whatever image is inputted.</p> |
