A video processing method includes determining first and second initial motion vectors of a current image block having a size of 64×64 pixels. The first and second initial motion vectors point to first and second reference images, respectively, which are forward and backward frames, respectively, of a current image containing the current image block. Distances from the first and second reference images to the current image are same. The method further includes performing or not performing a specified operation according to whether both the first and second reference images are short-term reference images determined according to sequence parameter set. The specified operation includes performing weighted summation on first and second reference image blocks pointed to by the first and second initial motion vectors, respectively, to obtain a weighted image block, and performing prediction for the current image block according to the weighted image block.