| 摘要 |
According to the present invention, an inter prediction encoding method comprises determining a motion vector and reference picture index of a current prediction unit; inducing a valid motion vector of prediction units at a pre-determined location adjacent to the current prediction unit to a space motion vector candidate; inducing a motion vector of a prediction unit, which is located at a corresponding location of or around the current prediction unit within a pre-encoded reference picture, to a valid time motion vector candidate of the current prediction unit; selecting either the valid space motion vector candidate or the valid time motion vector candidate as a motion vector predictor of the current prediction unit; calculating a motion vector difference value between the motion vector of the current prediction unit and the determined motion vector predictor; and encoding the difference value and the reference picture index. Therefore, the encoding quantity of motion information of the current prediction unit can be reduced, as the motion vector for existing images as well as an image that shows almost no motion or moves almost constantly in time and a background image can be expected by a motion vector candidate that can more effectively predict the motion. |
