An image encoding apparatus comprises a selector configured to select, from a set of candidate prediction operations each defining at least a prediction direction selected from a set of available prediction directions, a prediction operation for prediction of samples of a current region of a current image, the current region comprising an array of two or more rows and two or more columns of samples; and an intra-image predictor configured to derive predicted samples of the current region with respect to one or more of a group of reference samples of the same image in dependence upon a prediction direction, defined by the selected prediction operation, between a current sample to be predicted and the reference samples; in which for at least some of the candidate prediction operations: the group of reference samples comprises two or more parallel linear arrays of reference samples disposed at different respective separations from the current region, the intra-image predictor being configured to derive the predicted samples in dependence upon reference samples of the two or more linear arrays pointed to by the prediction direction; each of the linear arrays comprises reference samples at image positions, relative to the current region, within a horizontal and vertical image position range of the current region dependent upon a size and/or shape of the current region; the apparatus comprises a sample generator configured to generate zero or more additional reference samples for at least one of the linear arrays at one or more additional reference sample positions outside the horizontal and vertical image position range of the current region, so that, taking into account the additional reference samples, the prediction direction points to a reference position within each of the linear arrays.