obtaining, from a bitstream, a weight index;

obtaining a first prediction block by using a first reference image, and obtaining a second prediction block by using a second reference image, for bi-prediction of a current block;

obtaining a third prediction block by combining the first prediction block with the second prediction block according to a candidate value indicated by the weight index among candidate values included in a weight candidate group; and

reconstructing the current block by using the third prediction block,

wherein a first binary value corresponding to the weight index is entropy-decoded based on a context model,

wherein a remaining binary value corresponding to the weight index is entropy-decoded by a bypass method,

wherein a number of candidate values indicatable by the weight index in the weight candidate group when a picture order count (POC) of the first reference image and a POC of the second reference image are smaller than or equal to a POC of a current image is larger than a number of candidate values indicatable by the weight index in the weight candidate group when the POC of the first reference image and the POC of the second reference image are greater than the POC of the current image,

wherein the current image is split into a plurality of largest coding units, and

wherein a largest coding unit among the plurality of the largest coding units is hierarchically split into at least one coding unit including the current block.