Author Affiliations
1College of Artificial Intelligence, Yango University, Fuzhou, Fujian 350015, China2College of Physics and Information Engineering, Fuzhou University, Fuzhou, Fujian 350108, Chinashow less
Fig. 1. Schematic diagram of HS algorithm
Fig. 2. Decomposition diagram of NSCT
Fig. 3. Enhancement and contrast of coin images. (a) Original coin image; (b) image of low frequency sub-band; (c) processing result of low frequency sub-band; (d) image of high frequency scale 1; (e) image of high frequency scale 2; (f) image of high frequency scale 3; (g) processing result of high frequency scale 1; (h) processing result of high frequency scale 2; (i) processing result of high frequency scale 3; (j) enhancement result obtained by proposed method; (k) enhancement result obtained by homomorphic filtering method; (l) enhancement result obtained by wavelet transform; (m) enhancement result obtained by Retinex algorithm; (n) enhancement result obtained by method in Ref. [7]; (o) enhancement result obtained by method in Ref. [8]; (p) enhancement result obtained by method based on original QHS optimize to fuzzy sets
Fig. 4. Enhancement and contrast of human skeleton images. (a) Human skeleton image; (b) enhancement result obtained by proposed method; (c) enhancement result obtained by homomorphic filtering method; (d) enhancement result obtained by wavelet transform; (e) enhancement result obtained by Retinex algorithm; (f) enhancement result obtained by method in Ref. [7]; (g) enhancement result obtained by method in Ref. [8]; (h) enhancement result obtained by method based on original QHS optimize to fuzzy sets
Fig. 5. Enhancement and contrast of infrared images. (a) Infrared image; (b) enhancement result obtained by proposed method; (c) enhancement result obtained by homomorphic filtering method; (d) enhancement result obtained by wavelet transform; (e) enhancement result obtained by Retinex algorithm; (f) enhancement result obtained by method in Ref. [7]; (g) enhancement result obtained by method in Ref. [8]; (h) enhancement result obtained by method based on original QHS optimize to fuzzy sets
Fig. 6. Denoising and edge enhancement effect of coin noise images. (a) Original coin image; (b) noisy image; (c) enhancement result obtained by proposed method; (d) edge detection result of Fig. 6(c); (e) image of low frequency sub-band; (f) image of high frequency scale 1; (g) image of high frequency scale 2; (h) image of high frequency scale 3; (i) processing result of low frequency sub-band; (j) processing result of high frequency scale 1; (k) processing result of high frequency scale 2; (l) processing result of high frequency scale 3; (m) enhancement result obtained by homomorphic filtering method; (n) enhancement result obtained by wavelet transform; (o) enhancement result obtained by Retinex algorithm; (p) enhancement result obtained by method in Ref. [7]; (q) enhancement result obtained by method in Ref. [8]; (r) edge detection result of Fig. 6(a); (s) edge detection result of Fig. 6(m); (t) edge detection result of Fig. 6(n); (u) edge detection result of Fig. 6(o); (v) edge detection result of Fig. 6(p); (w) edge detection result of Fig. 6(q)
Function | Function formula | Parameter range of as | Optimal value |
---|
Rosenbrock | | [-10, 10] | 0 | Rastrigrin | | [-100, 100] | 0 | Schewefel | | [-500, 500] | 0 | Shubert | | [-100, 100] | 1 |
|
Table 1. Benchmark functions
Function | Optimal value of λ | Optimal value of Δθ |
---|
Rosenbrock | 1.5 | 0.4π | Rastrigrin | 1.7 | 0.4π | Schewefel | 1.7 | 0.5π | Shubert | 1.7 | 0.2π |
|
Table 2. Parameter testing of λ and Δθ
Algorithm | Average value | Optimal value | Worst value |
---|
HS | 2.04×10-1 | 1.82×10-2 | 5.60×10-1 | QGA | 1.56×10-3 | 2.99×10-5 | 1.40×10-2 | QHS | 3.22×10-5 | 2.15×10-9 | 2.10×10-4 | QBFA | 1.37×10-5 | 5.89×10-10 | 3.92×10-4 | Improved QHS | 5.44×10-6 | 7.21×10-11 | 1.18×10-4 |
|
Table 3. Performance testing of different algorithms
Method | Information entropy | Contrast ratio | Definition |
---|
Original image | 6.3153 | 18.4254 | 10.5878 | Homomorphic filtering | 6.2122 | 52.6358 | 21.1163 | Wavelet transform | 6.0037 | 21.1810 | 11.0198 | Retinex algorithm | 5.7908 | 41.0980 | 16.8536 | Method in Ref. [7] | 7.7662 | 17.1940 | 11.6188 | Method in Ref. [8] | 6.0610 | 37.2307 | 16.6904 | Method based on original QHS optimize to fuzzy sets | 6.4412 | 38.5640 | 18.5762 | Proposed method | 6.8128 | 43.0273 | 18.6284 |
|
Table 4. Quantitative comparison of enhancement effects of coin images
Method | Information entropy | Contrast ratio | Definition |
---|
Original image | 7.5286 | 7.0157 | 7.5282 | Homomorphic filtering | 6.9013 | 24.4868 | 16.5464 | Wavelet transform | 6.4962 | 9.4826 | 8.9901 | Retinex algorithm | 6.8555 | 15.4502 | 11.4918 | Method in Ref. [7] | 7.9618 | 7.1577 | 8.4743 | Method in Ref. [8] | 6.8649 | 16.1318 | 12.3101 | Method based on original QHS optimize to fuzzy sets | 6.6320 | 38.6521 | 11.5873 | Proposed method | 6.8128 | 43.0273 | 18.6284 |
|
Table 5. Quantitative comparison of enhancement effects of human skeleton images
Method | Information entropy | Contrast ratio | Definition |
---|
Original image | 5.9898 | 14.1011 | 8.8670 | Homomorphic filtering | 5.4445 | 45.2402 | 20.9698 | Wavelet transform | 6.2000 | 20.2866 | 11.6263 | Retinex algorithm | 5.5698 | 61.2892 | 25.0389 | Method in Ref. [7] | 7.0972 | 13.3551 | 9.2217 | Method in Ref. [8] | 4.1476 | 21.0801 | 10.6735 | Method based on original QHS optimize to fuzzy sets | 6.8268 | 40.0512 | 19.0563 | Proposed method | 6.9312 | 41.2877 | 19.9500 |
|
Table 6. Quantitative comparison of enhancement effects of infrared images
Method | PSNR /dB | Texture correlation |
---|
Homomorphic filtering | 7.1061 | 0.7746 | Wavelet transform | 10.1543 | 0.8655 | Retinex algorithm | 8.4954 | 0.7717 | Method in Ref. [7] | 10.4281 | 0.9254 | Method in Ref. [8] | 7.1022 | 0.9597 | Proposed method | 10.8431 | 0.9701 |
|
Table 7. PSNR and texture correlation of coin noisy images enhanced by different methods