An image processing apparatus is provided that allows accurate white balance gains to be obtained by accurately estimating the illuminant colors of various objects in an image. The white balance gains are obtained based on values of white pixels that have been extracted from an input image and have colors included in an extraction range, and the white balance gains are obtained based on specular reflection components included in the input image, are mixed in accordance with a mix ratio. The mix ratio is determined in accordance with, for example, a degree of reliability of extraction of the white pixels.