To increase the processing speed when median processing in the time direction is performed on color images, pixel values of each channel (R, G and B values) are extracted from multiple temporally consecutive frame images; subsequently, identifiers (index numbers) corresponding to frame images from which the pixel values are extracted are assigned to the pixel values; subsequently, a sorting process is performed on a set of pixel values of one channel (G values) to obtain an identifier assigned to a median value thereof; subsequently, a pixel value to which an identifier identical with that assigned to the median value is assigned is selected as a median value of each channel; and subsequently, a median image is generated on the basis of the median value of each channel obtained for each pixel.