Author Affiliations
1Faculty of Information Engineering and Automation, Kunming University of Science and Technology, Kunming, Yunnan 650500, China2College of Engineering & Science, Florida Institute of Technology, Melbourne, FL 32901, USAshow less
Fig. 1. Flow chart of fusion algorithm
Fig. 2. Images after pretreatment. (a) Filtered images; (b) binary images of filtered image; (c) differential images
Fig. 3. Double scale decomposition and mixing. (a)(b) Source image; (c) small-scale focusing diagram; (d) large-scale focusing diagram; (e) initial decision diagram
Fig. 4. Results of different r values processing
Fig. 5. Results of guided filtering and fusion. (a) Refined decision graph; (b) refined fusion graph; (c) final decision graph; (d) fusion result
Fig. 6. Images of part of common dataset used in experiment
Fig. 7. Decision graphs and fusion results obtained by different detection methods. (a) I1; (b) MWGF; (c) Ref. [24]; (d) CNN; (e) Ref. [27]; (f) GFDF; (g) proposed algorithm
Fig. 8. Residual images between I1 and fused image obtained by different algorithms. (a1)(a2) I1; (b1)(b2) NSCT_SR; (c1)(c2) MGFF; (d1)(d2) proposed algorithm
Fig. 9. Residual images between focused image and fused image obtained by proposed algorithm
Fig. 10. Influence of w on fusion index
Fig. 11. Influence of r on fusion index
Fig. 12. Influence of ε on fusion index
Algorithm | QMI | QNCIE | QM | QY | QCB |
---|
MWGF | 1.0090(2) | 0.8347(3) | 1.8367(2) | 0.9630(5) | 0.7787(0) | NSCT_SR | 0.8576(0) | 0.8262(0) | 1.3697(0) | 0.9384(0) | 0.7344(0) | CNN | 1.0276(2) | 0.8355(2) | 1.8118(0) | 0.9634(1) | 0.7874(2) | Ref. [24] | 1.0317(1) | 0.8359(1) | 1.8422(0) | 0.9629(6) | 0.7885(1) | MGFF | 0.7355(1) | 0.8207(1) | 0.5553(1) | 0.8639(0) | 0.6566(0) | GFDF | 1.0348(4) | 0.8359(4) | 1.8777(7) | 0.9633(4) | 0.7901(12) | Ref. [27] | 1.0348(8) | 0.8360(7) | 1.8801(5) | 0.9649(8) | 0.7896(5) | Proposed algorithm | 1.0358(9) | 0.8361(9) | 1.8829(12) | 0.9651(3) | 0.7893(7) |
|
Table 1. Comprehensive evaluation results of 27 groups of color images
Algorithm | QMI | QNCIE | QM | QY | QCB |
---|
MWGF | 0.9479(2) | 0.8273(2) | 1.1900(1) | 0.8632(3) | 0.7245(2) | NSCT_SR | 0.8661(0) | 0.8240(0) | 1.0160(0) | 0.8361(0) | 0.6951(0) | CNN | 1.0297(5) | 0.8281(6) | 1.4856(2) | 0.8949(4) | 0.7336(2) | Ref. [24] | 0.9567(2) | 0.8276(2) | 1.1898(0) | 0.8611(0) | 0.7260(0) | MGFF | 0.7616(0) | 0.8204(0) | 0.5516(0) | 0.7622(0) | 0.6225(0) | GFDF | 0.9565(1) | 0.8275(1) | 1.1998(1) | 0.8643(1) | 0.7298(4) | Ref. [27] | 0.9617(1) | 0.8277(0) | 1.2212(4) | 0.8637(2) | 0.7308(3) | Proposed algorithm | 0.9667(2) | 0.8321(2) | 1.2220(5) | 0.8740(3) | 0.7552(2) |
|
Table 2. Comprehensive evaluation results of 13 gray scale images
Image type(number) | Image size/(pixel×pixel) | MWGF | NSCT_SR | Ref. [24] | CNN | MGFF | GFDF | Ref. [27] | Proposedalgorithm |
---|
Gray(8) | 512×512640×480 | 4.20 | 159.48 | 4.17 | 154.09 | 0.80 | 0.48 | 2.13 | 0.37 | Color(15) | 520×520 | 9.83 | 166.14 | 3.85 | 131.22 | 2.21 | 0.46 | 2.01 | 0.49 |
|
Table 3. Average time calculated by different algorithms unit: s