Error concealment by means of motion refinement and regularized Bregman divergence.
Alessandra Martins CoelhoVania Vieira EstrelaFelipe P. do CarmoSandro R. FernandesPublished in: CoRR (2016)
Keyphrases
- error concealment
- bregman divergences
- motion vectors
- motion estimation
- packet loss
- information theoretic
- cost sensitive
- visual quality
- mahalanobis distance
- temporal error concealment
- image sequences
- motion field
- motion model
- maximum entropy
- macroblock
- moving objects
- exponential family
- optical flow
- loss function
- video quality
- nearest neighbor
- bitstream
- learning theory
- theoretical guarantees
- least squares
- kl divergence
- nonnegative matrix factorization
- bit rate
- video content
- machine learning
- high frequency
- maximum likelihood
- feature selection