Least squares motion estimation algorithm in the compressed DCT domain for H.26x/MPEG-x video sequences

A new compressed domain motion estimation algorithm that makes use of the DCT coefficients directly obtained from the H.26x or MPEG-x video stream is presented. The proposed algorithm is based on an iterative scheme that computes the new motion vectors by applying a least squares estimation technique. To reduce its computational effort, the algorithm may consider only an arbitrary subset of non- DCT coefficients. The performance of the algorithm was assessed in a DCT domain H.263 video transcoder, where the obtained motion vectors provided the means to significantly enhance the quality of the temporal prediction scheme with a consequent reduction of the required bit-rate.