| 摘要 |
PURPOSE:To obtain an image input device capable of accurately performing color correction even in an input state different from that when learning data is generated (difference in the diaphragm of the input device, that in the brightness of illumination color due to the difference of a light source, the secular change of a sensor, etc.). CONSTITUTION:A unit which handles information representing the state of input is added on a neural network by which relation between the color of an input image distorted by the color characteristic of the image input device and accurate color can be learned in addition to a unit which handles color information to be corrected, and a teacher signal 1 is generated based on the color information of a color chart 3 for learning to be inputted to the unit which handles the color information to be corrected. As information representing the state of input, specific single or plural reference color charts 2 are inputted simultaneously, and the color information of the reference color chart 2 can be used, or the mean value of the color information of the specific plural reference color charts 2 can be used, or the expression format of the color information of the reference color chart 2 is set at, for example, Lab color display format different from the expression format (additive color expression) of the color information to be corrected. |
