Author Affiliations
1Sino-German College of Intelligent Manufacturing, Shenzhen Technology University, Shenzhen, Guangdong 518118, China2Key Laboratory of Advanced Optical Precision Manufacturing Technology of Guangdong Provincial Higher Education Institute, Shenzhen, Guangdong 518118, China3Guangdong Provincial Key Laboratory of Micro/Nano Optomechatronics Engineering, Shenzhen, Guangdong 518118, Chinashow less
Fig. 1. Sharp and blurred images and their DWHT results. (a) Sharp image; (b) blurred image; (c) DWHT of Fig.1(a); (d) DWHT of Fig.1(b)
Fig. 2. Block diagram of proposed algorithm
Fig. 3. Focus region detection based on DWHT
Fig. 4. Generation process of final decision map. (a) Initial decision map IDM (x,y); (b) small patch removing operation MD1 (x,y); (c) morphology operation MD2 (x,y); (d) final decision map FDM (x,y)
Fig. 5. Three groups of multi-focus source images. (a) Clock; (b) book; (c) golf
Fig. 6. Clock’s fusion images and corresponding residual images of different methods. (a)(f) GFF; (b)(g) MWGF; (c)(h) CNN; (d)(i) DCT; (e)(j) proposed method
Fig. 7. Enlarged images of labeled regions in Fig. 6
Fig. 8. Multi-sequence focused images. (a) Multi-sequence gravure pattern appearance images; (b) multi-sequence lithium battery pole piece images
Fig. 9. Multi-focus fusion results of multi-sequence gravure pattern appearance image. (a) GFF; (b) CNN; (c) DCT; (d) proposed method
Fig. 10. Multi-focus fusion results of multi-sequence lithium battery pole piece image. (a) GFF; (b) CNN; (c) DCT; (d) proposed method
Fig. 11. Enlarged images of labeled regions in Fig. 10. (a) GFF; (b) CNN; (c) DCT; (d) proposed method
Fig. 12. Quality evaluation of multi-focus sequence image fusion
Evaluation metric | GFF | MWGF | CNN | DCT | Proposed method |
---|
QG | 0.6982 | 0.6958 | 0.7082 | 0.7064 | 0.7131 | QY | 0.7479 | 0.7806 | 0.7401 | 0.7443 | 0.7781 | QCB | 0.7584 | 0.7773 | 0.7821 | 0.7809 | 0.8357 | QFMI | 0.5862 | 0.5914 | 0.5958 | 0.5952 | 0.5980 |
|
Table 1. Average evaluation metrics of three groups of multi-focus source images under different methods