A progressively encoded image file, e.g. a JPEG bit stream, is decoded in multiple rounds. In first round, variable length encoded data in multiple scan segments of a first region are decoded. Meanwhile, position indicators for locating the next region are stored. In next round, the second region is decoded by reference to the position indicators for locating where the variable length encoded data of the second region are stored. The procedures are repeated until all regions are decoded to save memory usage during decoding.