摘要 |
The present invention relates to a method for processing a multi-viewpoint video based on a technique for interview prediction. A method for encoding a multi-viewpoint video according to an embodiment of the present invention comprises the steps of: generating depth information of the current block, which is an encoding target, based on a depth map of a video corresponding to a layer lower than that of the encoding target video at a first viewpoint; determining a disparity vector of the current block based on the depth information; extracting a block, which is included in the video at a second viewpoint, which is a base viewpoint, and is pointed by the disparity vector as a reference block; and encoding a signal indicative of a difference between the reference block and the current block. Furthermore, a method for decoding a multi-viewpoint video according to another embodiment of the present invention comprises the steps of: generating depth information of the current block, which is a decoding target, based on a depth map of a video which is included in a received bitstream and which corresponds to a layer lower than that of the decoding target video at a first viewpoint; determining a disparity vector of the current block based on the depth information; extracting a block, which is included in the video at a second viewpoint, which is a base viewpoint, and is pointed by the disparity vector as a reference block; and restoring the video by combining the difference signal relative to the reference block and the reference block. |