A method for predicting motion vectors over time based on sub-blocks, computers, and computer-readable storage media

Abstract

The invention relates to a method for predicting motion vectors over time based on a microcontroller, a computing device, and a computer-readable storage medium. The computing device collects a stream of video bits containing data combined with multiple encoded images. While decoding the current image in the video bitstream, the computing device, according to the syntactic components signaled in the video bitstream, selects a reference image as the same position as the current image, and determines the shift between the same-position image and the current image from the movement information of neighboring blocks in the spatial order of the current encoder unit (CU-Gude unit). Next, the computing device splits the current CU into several sub-CUs. The system obtains time-dependent motion vector predictions for each sub-unit from the displacement and motion information of the block in the image, along with the corresponding position of the sub-unit in the current image, and decodes the current sub-unit according to the time-dependent motion vector predictions.