Correction for color fog is securely achieved with limited load of processing. The RGB image data are converted into the luminance Y and the chromaticity values C1, C2, then a highlight point and a shadow point are determined and are connected to define the axis of the color solid. The color solid is rotated so as that the axis becomes parallel to the luminance axis Y and is so translated in parallel manner that the pixel of the minimum luminance coincides with the original point of the color space.