A video signal processing method comprises the steps of: acquiring reference samples on a first side of a current block and reference samples on a second side thereof on the basis of pre-restored samples neighboring the current block; acquiring a direct current (DC) value for a prediction of the current block on the basis of a reference sample set composed of at least some of the reference samples on the first side and the reference samples on the second side, wherein the reference sample set includes the number of reference samples raised to a power of 2, obtained by excluding some of the reference samples on the first side and the reference samples on the second side, if the length of the first side and the length of the second side are different; and restoring the current block on the basis of the DC value.