Infrared and Visible Image Fusion via Rolling Guidance Filtering and Hybrid Multi-Scale Decomposition

Cheng Zhao; Yongdong Huang

doi:10.3788/LOP56.141007

Journals >Laser & Optoelectronics Progress >Volume 56 >Issue 14 >Page 141007 > Article

Laser & Optoelectronics Progress
Vol. 56, Issue 14, 141007 (2019)

Infrared and Visible Image Fusion via Rolling Guidance Filtering and Hybrid Multi-Scale Decomposition

Cheng Zhao¹ and Yongdong Huang^1、2、*

Author Affiliations

¹ Institute of Image Processing and Understanding, North Minzu University, Yinchuan, Ningxia 750021, China

² Center of Mathematics and Information Science, Dalian Minzu University, Dalian, Liaoning 116600, China

show less

DOI: 10.3788/LOP56.141007 Cite this Article Set citation alerts

Cheng Zhao, Yongdong Huang. Infrared and Visible Image Fusion via Rolling Guidance Filtering and Hybrid Multi-Scale Decomposition[J]. Laser & Optoelectronics Progress, 2019, 56(14): 141007 Copy Citation Text

show less

Fig. 1. Basic framework of RGF

Download full size

Fig. 2. Smooth results of RGF corresponding to different σr values

Download full size

Fig. 3. Construction based on hybrid multi-scale decomposition

Download full size

Fig. 4. Five sets of infrared and visible images used in test. (a) Un camp; (b) Octec; (c) Steamboat; (d) Kayak; (e) Infrared4-Visible4

Download full size

Fig. 5. Fusion results obtained by different fusion methods applied in “Un camp” image

Download full size

Fig. 6. Fusion results obtained by different fusion methods applied in “Octec” image

Download full size

Fig. 7. Fusion results obtained by different fusion methods applied in “Steamboat” image

Download full size

Fig. 8. Fusion results obtained by different fusion methods applied in “Kayak” image

Download full size

Fig. 9. Fusion results obtained by different fusion methods applied in “Infrared4-Visible4” image

Download full size

Fig. 10. Statistical polygonal images of different objective evaluation indicators

Download full size

Fig. 11. Three-dimensional data images of “Octec”

Download full size

Fig. 12. Three-dimensional data images of “Infrared4-Visible4”

Download full size

σ_r	STD	SF	EI
0.1	0.1470	0.0305	0.0936
0.01	0.1514	0.0423	0.1290
0.001	0.1520	0.0443	0.1477
0.0001	0.1521	0.0443	0.1481
0.00001	0.1521	0.0443	0.1481
0.2	0.1457	0.0264	0.0909
0.02	0.1505	0.0399	0.1147
0.002	0.1520	0.0442	0.1466
0.0002	0.1521	0.0443	0.1481
0.00002	0.1521	0.0443	0.1481

Table 1. Objective evaluation indicators of visible image corresponding to different σr values

Image	Index	DWT	CVT	DTCWT	NSCT-DAPCNN	NSDTST	LATLRR	GTF	MSVD	ours
	MI	0.1916	0.1893	0.2011	0.2072	0.1802	0.2410	0.2793	0.2165	0.2841
	STD	0.1029	0.0988	0.0974	0.0718	0.1332	0.1070	0.1051	0.0892	0.1152 (2)
Un camp	Q_NCIE	0.8029	0.8029	0.8030	0.8030	0.8029	0.8034	0.8041	0.8031	0.8039 (2)
	Q_CV	425.1803	461.3446	426.053	423.8996	540.9333	332.3001	661.3364	424.1480	293.6759
	IE	0.4709	0.4685	0.4560	0.4580	0.6132	0.6268	0.6026	0.4500	0.8334
	MI	0.3876	0.3758	0.4127	0.4283	0.3838	0.4472	0.5176	0.4628	0.5200
	STD	0.1167	0.1175	0.1136	0.1117	0.2132	0.1175	0.0959	0.1080	0.1573 (2)
Octec	Q_NCIE	0.8066	0.8064	0.8070	0.8071	0.8073	0.8073	0.8089	0.8075	0.8094
	Q_CV	648.2625	408.1675	629.2341	626.6379	391.1703	274.6485	2264.0	1902.5	192.0751
	IE	0.9613	0.9645	0.9672	0.9660	0.9635	0.9934	0.8337	0.8298	1.0000
	MI	0.2576	0.2418	0.2695	0.2738	0.2911	0.2908	0.6284	0.2965	0.4514 (2)
	STD	0.0524	0.0482	0.0479	0.0449	0.124	0.0498	0.1331	0.0415	0.0769 (3)
Steamboat	Q_NCIE	0.8045	0.8043	0.8046	0.8046	0.8052	0.8047	0.8117	0.8047	0.8068 (2)
	Q_CV	163.4563	169.6701	167.6022	173.5677	298.9255	208.9697	331.4260	178.0732	138.259
	IE	0.3950	0.3742	0.3710	0.3567	0.6137	0.4527	0.5750	0.3516	0.7979
	MI	0.4471	0.3786	0.4492	0.5184	0.4923	0.4696	0.5469	0.6198	0.5195(3)
	STD	0.0731	0.0728	0.0696	0.0669	0.1707	0.0673	0.0864	0.0647	0.1444 (2)
Kayak	Q_NCIE	0.8059	0.8049	0.8058	0.8070	0.8070	0.8061	0.8072	0.8087	0.8087
	Q_CV	876.5883	622.103	864.7246	791.0015	798.335	855.5583	3060.6	697.9614	135.6981
	IE	0.0279	0.0224	0.0207	0.0132	0.7651	0.0262	0.0191	0.0071	0.9980
	MI	0.2485	0.2461	0.2866	0.2754	0.2552	0.2943	0.2390	0.2852	0.3233
	STD	0.1465	0.1366	0.1364	0.1289	0.1697	0.1432	0.1330	0.1243	0.1707
Infrared4-Visible4	Q_NCIE	0.8035	0.8034	0.8044	0.8038	0.8036	0.8040	0.8034	0.8039	0.8046
	Q_CV	1133.9	1056.2	1502.3	1076.9	1845.6	924.2383	1972.3	1045.83	718.6361
	IE	0.7643	0.7588	0.7511	0.7474	0.8973	0.8745	0.8450	0.7364	0.9504

Table 2. Objective evaluation indicators of different fusion methods

Images	DWT	CVT	DTCWT	NSCT-DAPCNN	NSDTST	LATLRR	GTF	MSVD	ours
Un camp	0.6707	2.1643	0.5408	41.7589	101.5510	227.6768	1.5259	0.1645	19.0259 (6)
Octec	0.2924	1.8312	0.4142	139.7033	326.4207	218.3082	7.8089	0.5087	58.0978 (6)
Steamboat	0.2856	1.8511	0.4191	118.7886	267.4719	251.7716	7.1331	0.4355	49.8169 (6)
Kayak	0.2777	1.8518	0.3950	117.9741	257.4367	245.7336	6.2976	0.6104	49.4241 (6)
Infrare4-Visible4	0.3054	1.8840	0.4115	27.8120	67.3537	220.7137	1.1274	0.2653	11.6429 (6)

Table 3. Running-time of eight other methods and proposed method

Cheng Zhao, Yongdong Huang. Infrared and Visible Image Fusion via Rolling Guidance Filtering and Hybrid Multi-Scale Decomposition[J]. Laser & Optoelectronics Progress, 2019, 56(14): 141007

Download Citation

Set citation alerts for the article

Tools

Set citation alerts for the article

Save the article for my favorites

Paper Information