Abstract | ||
---|---|---|
The computational time of HEVC encoder is increased mainly because of the hierarchical quad-tree-based structure, recursive coding units, and the exhaustive prediction search up to 35 modes. These advances improve the coding efficiency, but result in a very high computational complexity. Furthermore, selecting the optimal modes among all prediction modes is necessary for subsequent rate-distortion optimization process. Therefore, we propose a convolution neural network-based algorithm which learns the region-wise image features and performs a classification job. These classification results are later used in the encoder downstream systems for finding the optimal coding units in each of the tree blocks, and subsequently reduce the number of prediction modes. The experimental results show that our proposed learning-based algorithm reduces the encoder time saving up to 66.89% with a minimal Bjøntegaard delta bit rate (BD-BR) loss of 1.31% over the state-of-the-art machine learning approaches. Furthermore, our method also reduces the mode selection by 45.83% with respect to the HEVC baseline. |
Year | DOI | Venue |
---|---|---|
2019 | 10.1007/s00034-019-01110-4 | Circuits, Systems, and Signal Processing |
Keywords | Field | DocType |
CNN, Region of interest (RoI), CU partition, Angular mode selection, Softmax classifier | Mathematical optimization,Algorithmic efficiency,Convolutional neural network,Feature (computer vision),Algorithm,Coding (social sciences),Encoder,Artificial intelligence,Deep learning,Mathematics,Recursion,Computational complexity theory | Journal |
Volume | Issue | ISSN |
38 | 11 | 0278-081X |
Citations | PageRank | References |
9 | 0.62 | 0 |
Authors | ||
4 |
Name | Order | Citations | PageRank |
---|---|---|---|
Shiba Kuanar | 1 | 11 | 2.68 |
K. Raghunath Rao | 2 | 140 | 24.93 |
Monalisa Bilas | 3 | 9 | 0.62 |
Jonathan W. Bredow | 4 | 9 | 1.64 |