Adaptive CU Split Decision Based on Deep Learning and Multifeature Fusion for H.266/VVC

Table 5

The coding performance of the proposed algorithm compared with and previous works.

Test sequence		The proposed		FPIC [18]		ACSD [26]		FCPD [31]
		BD-rate (%)	TS (%)	BD-rate (%)	TS (%)	BD-rate (%)	TS (%)	BD-rate (%)	TS (%)
Class B; 1920 × 1080	Kimono	0.78	37.51	1.72	66.59	0.87	33.32	1.98	41.82
	ParkScene	0.61	39.56	1.28.	56.28	0.83	35.41	1.38	31.60
	BQTerrace	0.76	41.79	1.16	49.44	0.95	34.50	1.19	29.47
Class; C832 × 480	PartyScene	0.37	36.73	0.28	41.71	0.55	31.10	1.05	35.23
	RaceHorsesC	0.24	30.68	0.84	52.07	0.37	26.63	2.96	33.89
	BasketballDrill	1.25	39.21	1.91	53.05	1.30	33.39	1.36	28.73
Class; D416 × 240	BlowingBubbles	0.83	40.87	0.49	43.90	0.95	33.90	0.73	21.87
	RaceHorses	0.56	36.51	0.54	44.93	0.71	31.79	1.59	31.83
	BQSquare	0.58	36.67	0.17	32.34	0.68	30.73	-0.11	23.00
Class; E1280 × 720	Johnny	1.56	43.78	3.07	62.55	1.63	38.73	1.51	24.44
	FourPeople	1.34	46.51	2.55	62.18	1.38	38.01	1.37	26.65
	KristenAndSara	1.57	40.85	2.56	60.82	1.61	34.84	1.53	25.32
	Average	0.87	39.31	1.39	52.16	0.99	33.53	1.37	29.49