In the present embodiment, the tone of the input image data is 1 bit, the tone of the intermediate data is 1 bit, and the tone of the estimated image data is 8 bits. Note, however, that the technique of the present disclosure is not limited to these. In an example, the tone of the input image data may be 2 bits, the tone of the intermediate data may be 2 bits, and the tone of the estimated image data may be 16 bits.
The learning unit 205 includes a first learning unit 206, a second learning unit 207, and a learning intermediate data obtaining unit 208. The first learning unit 206 causes the first estimation unit 203 to learn the weights of layers in a neural network, which are its internal parameters. Likewise, the second learning unit 207 causes the second estimation unit 204 to learn the weights of layers in a neural network, which are its internal parameters.