In order to accurately determine a face area when performing image correction of an image, a pixel is set as a start point in the face area, then, pixels in which skin color is continuous from the start point are searched, and pixels that have been searched are linked to generate a skin color mask for the image. At the same time, an ellipse mask is generated for an ellipse area that includes the face area and is weighted according to distance from a center thereof, and which is composed with the skin color mask to generate a composition mask indicating the correction area.