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

Table 2

Partial unrolling algorithm on a single GPU unit. Execution times and speed-up values for several block size configurations and 3D datasets of normally distributed random numbers. Search and similarity windows have been set according to |𝑉_𝑖|=11³ and |_𝑗𝐵_𝑖|=3³.

Dataset size	Execution time/speed-up
	1 GPU unit				CPU
	(16, 16, 1)	(128, 1, 1)	(256, 1, 1)	(512, 1, 1)
64^3	0.99/17.7	0.77/22.7	1.32/13.3	2.72/6.43	17.5
128^2×64	2.31/35.9	2.25/36.9	2.23/37.2	4.57/18.2	83
128^3	4.62/36.8	4.5/37.8	4.46/38.1	9.22/18.4	170
256^2×128	16.9/41.2	16.2/42.9	16.6/42	18.6/37.3	696
256^3	33.79/41.2	32.5/42.9	33.14/42	37.5/37.2	1393
512^2×128	65.2/43.1	63.6/44.3	63.9/44	69.2/40.7	2814
512^2×256	131/43.1	128/43.9	128/43.8	139/40.5	5623
512^3	264/42.7	261/43.3	259/43.5	281/40.2	11291