Image Fusion Based on Residual Learning and Visual Saliency Mapping

Luoyi Ding; Jin Duan; Yu Song; Yong Zhu; Xiaoshan Yang

doi:10.3788/LOP57.161008

Journals >Laser & Optoelectronics Progress >Volume 57 >Issue 16 >Page 161008 > Article

Laser & Optoelectronics Progress
Vol. 57, Issue 16, 161008 (2020)

Image Fusion Based on Residual Learning and Visual Saliency Mapping

Luoyi Ding¹, Jin Duan^1、2、*, Yu Song¹, Yong Zhu³, and Xiaoshan Yang¹

Author Affiliations

¹College of Electronic Information Engineering, Changchun University of Science and Technology, Changchun, Jilin 130022, China

²Fundamental Science on Space-Ground Laser Communication Technology Laboratory, Changchun University of Science and Technology, Changchun, Jilin 130044, China

³College of Computer Science and Technology, Changchun University of Science and Technology, Changchun, Jilin 130022, China

show less

DOI: 10.3788/LOP57.161008 Cite this Article Set citation alerts

Luoyi Ding, Jin Duan, Yu Song, Yong Zhu, Xiaoshan Yang. Image Fusion Based on Residual Learning and Visual Saliency Mapping[J]. Laser & Optoelectronics Progress, 2020, 57(16): 161008 Copy Citation Text

show less

Fig. 1. Fusion framework of visible light image and infrared image

Download full size

Fig. 2. Mapping relationship between α and weighting coefficient

Download full size

Fig. 3. Deep learning fusion framework for edge layer and texture layer

Download full size

Fig. 4. Fusion results of Steamboat source images by different methods. (a) Infrared image; (b) visible light image; (c) CBF; (d) WLS; (e) HMSD; (f) JSR; (g) JSRD; (h) CSR; (i) proposed method

Download full size

Fig. 5. Fusion results of Street source images by different methods. (a) Infrared image; (b) visible light image; (c) CBF; (d) WLS; (e) HMSD; (f) JSR; (g) JSRD; (h) CSR; (i) proposed method

Download full size

Fig. 6. Fusion results of Marne source images by different methods. (a) Infrared image; (b) visible light image; (c) CBF; (d) WLS; (e) HMSD; (f) JSR; (g) JSRD; (h) CSR; (i) proposed method

Download full size

Picture name	Metric	CBF	WLS	HMSD	JSR	JSRD	CSR	Proposed method
Steamboat	FMIdct	0.2192	0.3446	0.3379	0.1795	0.1489	0.3547	0.3879
	FMIw	0.2462	0.3636	0.3415	0.1812	0.1483	0.3044	0.3773
	Nabf	0.5448	0.1938	0.1583	0.2552	0.4102	0.0207	0.0139
	SSIM	0.5828	0.8577	0.8382	0.7420	0.6894	0.8751	0.8836
Street	FMIdct	0.2846	0.3492	0.3574	0.2485	0.2279	0.3760	0.3673
	FMIw	0.2486	0.3575	0.3453	0.2769	0.2642	0.3220	0.3802
	Nabf	0.4870	0.1352	0.1299	0.1804	0.1908	0.0220	0.0190
	SSIM	0.4986	0.6709	0.6565	0.6299	0.6237	0.6747	0.6854
Marne	FMIdct	0.1835	0.3236	0.3092	0.1591	0.1324	0.2803	0.3569
	FMIw	0.2407	0.3773	0.3416	0.2266	0.2078	0.3441	0.4254
	Nabf	0.5278	0.1150	0.1636	0.2109	0.2801	0.0193	0.0155
	SSIM	0.4575	0.6769	0.6956	0.6253	0.5808	0.7218	0.7199

Table 1. FMIdct, FMIw, Nabf, and SSIM values of the fused images Steamboat, Street, and Marne

Metric	CBF	WLS	HMSD	JSR	JSRD	CSR	Proposed method
FMIdct	0.2631	0.3349	0.3319	0.1644	0.1444	0.3464	0.3762
FMIw	0.3235	0.3803	0.3690	0.2083	0.1842	0.3836	0.4183
Nabf	0.3173	0.2232	0.1543	0.2393	0.3510	0.0196	0.0162
SSIM	0.5996	0.7194	0.7221	0.6064	0.5410	0.7422	0.7714

Table 2. Mean values of FMIdct, FMIw, Nabf, and SSIM for twenty-one pairs of images

Luoyi Ding, Jin Duan, Yu Song, Yong Zhu, Xiaoshan Yang. Image Fusion Based on Residual Learning and Visual Saliency Mapping[J]. Laser & Optoelectronics Progress, 2020, 57(16): 161008

Download Citation

Set citation alerts for the article

Tools

Set citation alerts for the article

Save the article for my favorites

Paper Information