A video comprises at least one shot (SH), which is a continuous sequence of images representing a scene viewed from a particular location. Images are selected from a shot (SH) so as to obtain a continuous sequence of selected images (SI) that are evenly distributed throughout the shot. At least one continuous subsequence (SB1, SB2, SB3) of selected images that meet a predefined similarity test is identified. An image is selected from a continuous portion (SP) of the shot that coincides in time with the longest continuous subsequence (SB2) of selected images that meet the predefined similarity test. The image that is selected constitutes a representative image (RI) for the shot.