The first estimation unit 203 learns a model that receives input image data as shown in FIG. 3A and outputs intermediate data as shown in FIG. 3C. The model that converts the 1-bit per pixel input image data shown in FIG. 3A into the 1-bit per pixel intermediate data in FIG. 3C contains random edge changes originating from the scattering or bleeding of the color material. Here, the intermediate data outputted from the first estimation unit 203 is binary image data and does not have tones to express the gradations of the color material. Hence, the estimation by the first estimation unit 203 can be said to include no estimation of the blurs in the image originating from the gradations of the color material.
The second estimation unit 204 learns a model that receives intermediate data as shown in FIG. 3C and estimates scanned image data of a printing result having an 8-bit tone per pixel as shown in FIG. 3B. In FIGS. 3B and 3C, the pixel groups with the lowest (black) pixel value have the same outline, and the edges are therefore not changed. That is, the estimation by the second estimation unit 204 hardly includes an estimation of the random edge changes originating from the scattering or bleeding of the color material. This enables image conversion from binary intermediate data into a multi-value scanned image that reproduces blurs equivalent to those in the actual image.
<Configuration of First Estimation Unit 203>
