Low complexity video encoding with one-bit transform based network-driven motion estimation

Abstract

Low complexity video encoding is desired in many applications requiring real-time encoding of video, particularly for mobile equipment. In order to reduce encoding complexity without much sacrifice in visual quality, some of the complexity can typically be shifted to the decoder, or the network. infrastructure can take part in the process by performing some encoding and/or decoding tasks. In this paper a new hybrid video coding scheme is proposed that utilizes the standard motion compensated predictive coding architecture and uses one-bit transform (1BT) representations of video frames to facilitate remote motion estimation, The one-bit transform of each video frame is computed at the encoder using low complexity operations and sent to the decoder or network infrastructure, after entropy encoding, for motion estimation. After motion estimation is carried out remotely, the motion vectors are sent back to the encoder for motion compensated predictive coding. In order to reduce the overhead introduced by one-bit transforms, differential encoding of 1BTs is also investigated. The proposed approach can provide low complexity motion compensated predictive video encoding by shifting the high computational load of motion estimation to the decoder or network infrastructure(1).