Back to Search
Document details
CE5: Per-context CABAC initialization with double windows (Test 5.1.3)
Abstract
This contribution reports the simulation results for Test CE5.1.3 [1]. In this test, the CABAC engine uses double windows to estimate the binary symbol probabilities, where the window parameters are different for different contexts and are stored in a pre-defined look-up table (LUT). The range update method used for the CABAC engine is either the one proposed in JVET-L0117 [2] or the one proposed in JVET-L0335 [3]. Compared with VTM3.0, the average coding gains of the proposed CABAC engine are -0.99% AI, -0.95% RA, and -0.82% LDB, when JVET-L0117 is used for range update; and are -0.95% AI, -0.90% RA, and -0.77% LDB, when JVET-L0335 is used for range update.
JVET-M0412 CE5: Per-context CABAC initialization with double windows (Test 5.1.3) [A. Said, J. Dong, H. Egilmez, Y.-H. Chao, M. Karczewicz, V. Seregin (Qualcomm)]
References:
Cited:
PATENTS:
ES2720662T3
0.34
2019-07-23
US12537938B2
0.32
2026-01-27
US11871025B2
0.30
2024-01-09
US11695946B2
0.28
2023-07-04
US20240195982A1
0.26
2024-06-13
CN113261292A
0.24
2021-08-13
CN119137944A
0.22
2024-12-13
CN119563315A
0.20
2025-03-04
Decisions
JVET-M0412 CE5: Per-context CABAC initialization with double windows (Test 5.1.3) [A. Said, J. Dong, H. Egilmez, Y.-H. Chao, M. Karczewicz, V. Seregin (Qualcomm)]
Citation