Inter-prediction methods based on linear embedding for video compression

This paper considers the problem of temporal prediction for inter-frame coding of video sequences using locally linear embedding (LLE). LLE-based prediction, first considered for intra-frame prediction, computes the predictor as a linear combination of K nearest neighbors (K-NN) searched within one or several reference frames. The paper explores different K-NN search strategies in the context of temporal prediction, leading to several temporal predictor variants. The proposed methods are tested as extra inter-frame prediction modes in an H.264 codec, but the proposed concepts are still valid in HEVC. The results show significant Rate-Distortion performance gains are obtained with respect to H.264 (up to 15.31 % bit-rate saving).