Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Line-based intra prediction for next-generation video coding

: De-Luxan-Hernandez, S.; Schwarz, H.; Marpe, D.; Wiegand, T.


Institute of Electrical and Electronics Engineers -IEEE-:
25th IEEE International Conference on Image Processing, ICIP 2018 : October 7-10, 2018, Athens, Greece
Piscataway, NJ: IEEE, 2018
ISBN: 978-1-4799-7061-2
ISBN: 978-1-4799-7062-9
International Conference on Image Processing (ICIP) <25, 2018, Athens>
Conference Paper
Fraunhofer HHI ()

Intra prediction in modern block-based hybrid video coding standards like H.265/High Efficiency Video Coding (HEVC) extrapolates the reconstructed values of neighboring reference samples of a block to generate the prediction signal. Due to the loss of correlation between samples with increasing distance, this procedure may lead to residual signals with high levels of energy concentrated in the most distant regions of the block relative to the neighboring reference samples. To circumvent this loss in prediction quality with increasing distance and improve the coding efficiency of intra prediction, this publication proposes an algorithm that divides blocks into line-based sub-partitions that are processed individually. This approach has been implemented in the context of HEVC and the new Joint Exploration Test Model (JEM), with the result of obtaining average bit-rate savings of 2.77% and 1.62% for a whole test set of video sequences according to the All Intra and Random Access configurations, respectively.