3D Data Denoising via Nonlocal Means Filter by Using Parallel GPU Strategies

Table 10

Full unrolling algorithm with split for different video formats. Observe that the number of splits is intended on the same GPU and the time needed to copy data from CPU to GPU and from GPU to CPU is reported in bold.

Video size	Splits	Execution time + CPU-GPU copy time
		(11^3,3^3)	(21^3,3^3)	(11^3,5^3)	(21^3,5^3)
PAL ×50	1	16.8 + 0.1	85.8 + 0.15	55 + 0.11	276 + 0.16
PAL ×100	2	33.5 + 0.2	172 + 0.3	110 + 0.22	552 + 0.32
PAL ×200	4	67 + 0.4	343 + 0.59	220 + 0.43	1104 + 0.63
PAL ×500	10	168 + 0.99	858 + 1.48	550 + 1.08	2761 + 1.58
HD ×50	1	31.4 + 0.18	161 + 0.27	103 + 0.2	517 + 0.29
HD ×100	2	62.8 + 0.37	321 + 0.54	206 + 0.4	1035 + 0.58
HD ×200	4	126 + 0.73	643 + 1.09	412 + 0.8	2070 + 1.16
HD ×500	10	314 + 1.83	1607 + 2	1030 + 4.8	5175 + 2.9
FULL-HD ×50	1	78.5 + 0.61	402 + 1.04	257 + 0.64	1294 + 1.1
FULL-HD ×100	2	157 + 1.23	804 + 2.07	515 + 1.28	2587 +2.2
FULL-HD ×200	4	314 + 2.45	1607 + 4.14	1030 + 2.56	5174 + 4.4
FULL-HD ×500	10	785 + 6.14	4018 + 10.4	2575 + 6.4	12935 + 11