Nonrigid Image Registration in Digital Subtraction Angiography Using Multilevel B-Spline

Table 3

Comparison of final registration results of the iterative refinement algorithm [19], the multiresolution blocking algorithm [20], and the proposed algorithm on ten simulated image pairs.

Image pairs	Initial error before registration	After registration
		Iterative refinement			Multiresolution blocking			Proposed approach
	RMS (𝛿0)	RMS	(𝛿0−𝛿)/𝛿0%	Elapsed time (s)	RMS	(𝛿0−𝛿)/𝛿0%	Elapsed time (s)	RMS	(𝛿0−𝛿)/𝛿0%	Elapsed time (s)
No. 1	6.098	1.942	68.1%	268.8	1.511	75.2%	8954	1.486	75.6%	14.9
No. 2	6.892	1.396	79.7%	270.5	1.015	85.3%	9958	0.881	87.2%	13.5
No. 3	6.442	1.682	73.9%	269.8	0.870	86.5%	8745	1.302	79.8%	13.1
No. 4	6.201	1.191	80.8%	274.5	1.603	74.1%	9180	1.061	82.9%	15.6
No. 5	6.363	1.825	71.3%	254.8	1.254	80.3%	9553	1.289	79.8%	14.4
No. 6	6.270	1.258	79.9%	275.1	1.542	75.4%	9210	0.960	84.7%	12.4
No. 7	6.813	2.133	68.7%	296.3	1.532	77.5%	8768	1.614	76.3%	14.2
No. 8	6.920	1.349	80.5%	265.0	1.272	81.6%	9003	1.229	82.2%	14.3
No. 9	7.399	1.846	75.1%	289.6	1.784	75.9%	9548	1.490	79.9%	14.2
No. 10	7.494	2.408	67.9%	256.5	2.112	71.8%	9346	1.643	78.1%	15.1
Average	—	—	74.6%	272.1	—	78.4%	9226	—	80.6%	14.2