A method, an apparatus and computer programs are disclosed for forming synthesized panning of images. Based on a pair of digital images taken by a pair of digital cameras and their calibration information, a disparity map is formed for image objects in the pair of digital images. A combined image is formed using the pair of digital images. The combined image is segmented, using the disparity map, to comprise a foreground region and a background region. A sequence of synthesized panning images is formed so that for each combined image: a perspective shift is applied between the foreground region and background region; and a shifting portion of the perspective shifted image is cropped.