FIG. 11 shows an example configuration of the learning unit 903. The image estimation processing unit 902, which is the learning target, calculates an estimated image as scanned image data estimated from input image data.

A determination unit 1101 determines whether inputted image data is the correct data or the estimated image data outputted by the image estimation processing unit 902. The determination unit 1101 returns 1 if determining that the inputted image data is the correct data, and returns 0 if determining that the inputted image data is the estimated image data outputted by the image estimation processing unit 902. A configuration of the determination unit 1101 will be described later.

An error calculation unit 1102 calculates an error based on the result of the determination by the determination unit 1101 on the estimated image and the correct image. A parameter updating unit 1103 updates the parameters of the image estimation processing unit 902 and the determination unit 901 based on the error calculated by the error calculation unit 1102. Note that the process in the learning unit 902 in the present embodiment is similar to that of the first learning unit 201 in embodiment 1.

<Configuration of Determination Unit 1101>

A configuration of the determination unit 1101 will be described below. FIG. 12 shows an example configuration of the neural network in the determination unit 1101. Although the process in the present embodiment will be described below based on the neural network shown in FIG. 12, the technique of the present disclosure is not limited to it.