Back to Search
Document details
Pending archive for JVET-N0000
Abstract
This contribution presents a modified version of the Decoder-side Intra Mode Derivation (DIMD) on top of the previously proposed contribution JVET-M0094 [1]. The proposed DIMD skips the intra mode transmission and performs a texture gradient processing to derive the best mode, instead. Compared to the previous version, the main modification is a “prediction fusion” step which uses up to three intra modes, implicitly derived from the texture. Like the previous version, the second and the third intra modes for prediction fusion are also derived by the texture gradient processing step. The results show a 0.41% gain on average in AI for 125% encoding time and 110% decoding time.
References:
PATENTS:
US9484077B2
0.34
2016-11-01
CN106559669B
0.32
2018-10-09
US20190304102A1
0.30
2019-10-03
US11587432B2
0.28
2023-02-21
US8755837B2
0.26
2014-06-17
TWI827606B
0.24
2024-01-01
US9697512B2
0.22
2017-07-04
JP6042470B2
0.20
2016-12-14
Citation