An apparatus for generating a three-dimensional image representation of a scene comprises a receiver (301) receiving a tiled three-dimensional image representation of a scene from a first viewpoint, the representation comprising a plurality of interconnected tiles, each tile comprising a depth map and a texture map representing a viewport of the scene from the first viewpoint and the tiles forming a tiling pattern. A first processor (311) determines neighboring border regions in at least a first tile and in a second tile in response to the tiling pattern. A second processor (309) modifies at least a first depth value of a first border region of the first tile in response to at least a second depth value in a second border region of the second tile, the border regions being neighboring regions. The modified depth maps may be used to generate a mesh based on which images may be generated.