An image processing apparatus that can be applied to a wide range of general images and that is capable of improving shitsukan of materials depicted in an image by interactively adjusting multiple physical image characteristics of an image based on human latent sensations. Physical image-characteristic parameters are extracted from an input image. The image-characteristic parameters are processed using a first numerical conversion equation, and various low-order-sensitivity parameters quantifying general latent sensations are calculated. The low-order-sensitivity parameters are processed using a second numerical conversion equation, and high-order-sensitivity parameters quantifying shitsukan for different materials are calculated. An output image of which the shitsukan in the input image are improved by combining various types of image processing for changing the image characteristics so that the high-order-sensitivity parameters approach predetermined values is generated.