To encode an image divided into blocks of a set of images, each block being encoded according to one of a plurality of encoding modes including at least one temporal correlation prediction encoding mode utilizing a plurality of images, a motion estimation vector search area is defined in a second image of the set of images, distinct from the first image and previously encoded according to a predefined sequence of encoding images of the set of images, a portion at least of the search area having substantially the shape of an ovoid, the data from the search area is stored in a cache memory, a motion estimation vector of the current block is determined by a search in the search area in the cache memory, and the motion estimation vector is used to decide the encoding of the current block according to the one of the plurality of encoding modes.