Motion vector composition algorithm for spatial scalability in compressed video

Spatial scalable encoding of visual data has several applications, including browsing visual databases, querying multimedia databases, interactive multimedia communications, etc. To convert a compressed video sequence to a lower spatial resolution compressed video, it may decode the original bit-stream, down-sample each frame, and encode. The computation-intensive motion estimation will be included in the transcoder. The re-use of motion vectors extracted from incoming video bit-streams during transcoding has been widely accepted. This paper proposes a new algorithm to estimate the motion vector of the reduced low-resolution video by only using the original motion vectors and DC coefficients in the compressed bit-stream. By taking into account the spatial activity measurement to generate better prediction for the composed motion vector, higher quality and lower complexity than previous works are achieved by the proposed method.