Efficient motion estimation in hierarchical structure

摘要

Techniques for coding of video content are presented. An encoder component generates a down-sampled image(s) of a video frame of a video frame sequence to form a multi-resolution level video structure. Using a fast motion estimation algorithm, the encoder component performs motion estimation on a lowest-resolution image, propagates estimation results to the next-higher resolution level. The encoder component uses those estimation results, and/or spatial information of previous encoded macroblocks of the next-higher resolution image, as a prediction for motion estimation performed on the next-higher resolution image. The encoder component applies mode selection and motion search rules, and calculates and uses dissimilarity values of macroblocks, to facilitate prediction for motion estimation at the next-higher resolution level. The encoder component can continue this motion estimation process for another higher-resolution image(s) of the structure up through the original image.