Author Affiliations
1School of Optics and Photonics, Beijing Institute of Technology, Beijing 100081, China2School of Integrated Circuits and Electronics, Beijing Institute of Technology, Beijing 100081, Chinashow less
Fig. 1. Schematic diagram of underwater formation model
Fig. 2. Flowchart of HSV-CIELAB classification equalization and minimum convolution area DCP
Fig. 3. CIELAB color equalization results. (a) RGB channel grayscale of raw image; (b) RGB channel grayscale of balanced image
Fig. 4. Comparison of LAB histograms of images. (a) Channel L of raw image; (b) channel a of raw image; (c) channel b of raw image; (d) channel L of balanced image; (e) channel a of balanced image; (f) channel b of balanced image
Fig. 5. Composite false color image
Fig. 6. Backlight estimation. (a) Original image; (b) dark channel ; (c) convolution image
Fig. 7. Example of backlight estimation with minimum convolution area DCP algorithm
Fig. 8. Comparison of visual results of different algorithms. (a) (b) High-saturation distorted images; (c) (d) low-saturation distortion images; (e) (f) shallow water images
Fig. 9. Enhancement results of different algorithms and comparison with reference images. (a) (b) (c) High-saturation distorted images; (d) (e) (f) (g) (h) low-saturation distorted images; (i) shallow water image
Algorithm | Parameter | UDCP | ULAP | IBLA | MIP | Proposed algorithm |
---|
Image(a) | PSNR | 10.8911 | 18.8947 | 18.7877 | 15.4244 | 19.1593 | SSIM | 0.4892 | 0.7377 | 0.7694 | 0.6969 | 0.7298 | Image(b) | PSNR | 12.9269 | 17.5037 | 20.7304 | 19.3203 | 21.1145 | SSIM | 0.6779 | 0.7033 | 0.8295 | 0.7530 | 0.8716 | Image(c) | PSNR | 12.5384 | 16.5213 | 17.4517 | 19.1754 | 21.9297 | SSIM | 0.6335 | 0.6913 | 0.6356 | 0.8048 | 0.8136 | Image(d) | PSNR | 10.3364 | 17.0962 | 19.3123 | 16.6283 | 21.8104 | SSIM | 0.6046 | 0.7937 | 0.8463 | 0.7732 | 0.8943 | Image(e) | PSNR | 15.7936 | 18.8630 | 17.6148 | 17.7478 | 22.7959 | SSIM | 0.5197 | 0.8053 | 0.8704 | 0.8502 | 0.8880 | Image(f) | PSNR | 15.8219 | 17.3456 | 16.4539 | 16.7591 | 18.3918 | SSIM | 0.7315 | 0.8510 | 0.8077 | 0.7959 | 0.8590 | Image(g) | PSNR | 11.1105 | 15.9795 | 16.3614 | 16.1424 | 20.3412 | SSIM | 0.6959 | 0.8365 | 0.8351 | 0.8262 | 0.8878 | Image(h) | PSNR | 18.0405 | 21.3973 | 20.7642 | 21.1455 | 22.7646 | SSIM | 0.8391 | 0.9116 | 0.9283 | 0.9338 | 0.9188 | Image(i) | PSNR | 13.5442 | 13.6196 | 15.8227 | 13.5737 | 18.6227 | SSIM | 0.7319 | 0.7633 | 0.6139 | 0.7566 | 0.8931 |
|
Table 1. PSNR and SSIM evaluation quality of different algorithms
Algorithm | UDCP | ULAP | IBLA | MIP | Proposed algorithm | Reference |
---|
Image(a) | 0.4629 | 0.5340 | 0.5622 | 0.5275 | 0.5710 | 0.5777 | Image(b) | 0.5751 | 0.5827 | 0.6099 | 0.6105 | 0.5934 | 0.5778 | Image(c) | 0.4869 | 0.5527 | 0.4744 | 0.5451 | 0.5756 | 0.5718 | Image(d) | 0.4592 | 0.5712 | 0.5502 | 0.5109 | 0.5746 | 0.5581 | Image(e) | 0.5319 | 0.5738 | 0.5912 | 0.5887 | 0.5829 | 0.6139 | Image(f) | 0.5386 | 0.5849 | 0.5906 | 0.5545 | 0.5837 | 0.5031 | Image(g) | 0.5570 | 0.5817 | 0.5864 | 0.5898 | 0.6081 | 0.6074 | Image(h) | 0.5818 | 0.6022 | 0.5680 | 0.5862 | 0.5685 | 0.5670 | Image(i) | 0.5244 | 0.5465 | 0.5315 | 0.5498 | 0.5541 | 0.5287 |
|
Table 2. Parameters of UCIQE of different algorithms