Image retargeting is appropriately performed on stereo pair images composed of at least two images such as in three-dimensional displays. A path of connected pixels in first image data is calculated based on pixel gradient energy. Each pixel in second image data corresponding to each pixel in connected pixels in the first image data is calculated as an initial search point, based on the stereo correspondence relationship between the first image data and the second image data. Pixels that minimize energy between pixels of the first image data and pixels of the second image data in the proximity of the initial search point is calculated as a path of connected pixels in the second image data. A path of optimal connected pixels in the first image data is calculated using the energy.