摘要 |
PROBLEM TO BE SOLVED: To attain image magnification coping flexibly with various received images. SOLUTION: A neural network of a pre-stage needs 9, 9, d sets of neurons respectively and a post-stage needs (9, 9, d)×d sets of neurons. Then d-sets of neurons configuring an output layer 3a of the neural network of the pre-stage correspond one to one to d-sets of networks of the neural network of the post- stag. The neural network of the pre-stage provides an output of consecutive values 1-0 in each of the d-sets of neurons configuring the output layer 3a, and a neuron having a maximum value in this case is used for an ignition neuron. Then the network of the post-stage corresponding to the ignition neuron is used and an output of the network of the post-stage is selected for an upper left pixel y1(2i, 2j). |