1. Single Image Haze Removal Using Dark Channel Prior
Kaiming He Jian Sun Xiaoou Tang
The Chinese University of Hong Kong Microsoft Research Asia
The Chinese University of Hong Kong

2. Hazy Images
Low visibility
Faint colors

3. Goals of Haze Removal
depth
Scene restoration
Depth estimation

4. Haze Imaging Model I  J  t  A  (1 t) Atmospheric light Hazy image
Scene radiance
Transmission

5. Haze Imaging Model I  J  t  A  (1 t) d   ln t Depth
Transmission

6. Ambiguity in Haze Removal
scene
radiance ….
input
depth

7. Previous Works
Single image
– Maximize local contrast
[Tan, CVPR 08]

8. Previous Works
Single image
– Maximize local contrast
[Tan, CVPR 08]

9. Previous Works Single image – Maximize local contrast
[Tan, CVPR 08]
– Independent Component Analysis [Fattal, Siggraph 08]

10. Previous Works Single image – Maximize local contrast
[Tan, CVPR 08]
– Independent Component Analysis [Fattal, Siggraph 08]

11. Priors in Computer Vision
Ill-posed problem
well-posed problem
Smoothness prior •
Sparseness prior Exemplar-based prior
Dark Channel Prior

12. Dark Channel
min (rgb, local patch)

13. Dark Channel
min (rgb, local patch)
– min (r, g, b)
min (r, g, b)

14. Dark Channel min (rgb, local patch) min (r, g, b)
min (local patch) = min filter
15 x15
dark channel
darkest

15. Dark Channel Jdark (x)  min ( min Jc (y)) min (rgb, local patch)
min (local patch) = min filter
Jdark (x)  min ( min Jc (y))
y(x) c{r,g,b}
Jc: color channel of J
Jdark: dark channel of J
dark channel

16. Dark Channel Jdark  min (min Jc ) min (rgb, local patch)
min (local patch) = min filter
Jdark  min (min Jc )
 c
Jc: color channel of J
Jdark: dark channel of J
dark channel

17. A Surprising Observation
Haze-free

18. A Surprising Observation
Haze-free

19. A Surprising Observation
Haze-free

20. A Surprising Observation
Haze-free

21. A Surprising Observation
Haze-free

22. A Surprising Observation
Haze-free

23. A Surprising Observation1
Prob.
0.8
86% pixels
in [0, 16]
0.6
5,000 haze-free images
0.4
0.2
Pixel intensity of dark channels
256

24. Dark Channel Prior
For outdoor haze-free images
min (min Jc )  0
 c

25. What makes it dark?
Shadow
Colorful object
Black object

26. Dark Channel of Hazy Image
hazy image dark channel
The dark channel is no longer dark.

27. Transmission Estimation
Haze imaging model
I  J  t  A  (1 t)
I  J c c
Normalize
t 1 t
Ac Ac
Compute dark channel Ic Jc  
min (min
 c
)  min (min c
)t 1 t c  A A 
 c

28. Transmission Estimation
Dark Channel Prior
min (min Jc )  0
 c
Compute dark channel
 0 Ic Jc  
min (min
 c
)  min (min c
)t 1 t c  A A 
 c

29. Transmission Estimation
Estimate transmission c I
t  1 min (min ) Ac
 c
Compute dark channel
 Jc 
min (min c )t 
  c A  I c
min (min
 c
) 
1 t Ac

30. Transmission Estimation
Estimate transmission c I
t  1 min (min
 c ) Ac
estimated t
input I

31. Transmission Optimization
Haze imaging model I  J  t  A  (1 t)
I  F   B  (1 )
Matting model + 
tri-map
Refined transmission +

32. Transmission Optimization
~ 2
(t)   t  t
tTLt
Data term
Smoothness term
L - matting Laplacian [Levin et al., CVPR ‘06]
Constraint - soft, dense (matting - hard, sparse)

33. Transmission Optimization
before optimization

34. Transmission Optimization
after optimization

35. Atmospheric Light Estimation
A: most hazy
brightest pixels
brightest pixel
hazy image
dark channel

36. Scene Radiance Restoration
Atmospheric light
I  J  t  A  (1 t)
Hazy image
Scene radiance
Transmission

37. Results
input

38. Results
recovered image

39. Results
depth

40. Results
input

41. Results
recovered image

42. Results
depth

43. Results
input

44. Results
recovered image

45. Results
depth

46. Comparisons
input
[Fattal Siggraph 08]

47. Comparisons
input
our result

48. Comparisons
input
[Tan, CVPR 08]

49. Comparisons
input
our result

50. Comparisons
input
[Kopf et al, Siggraph Asia 08]
our result

51. Results: De-focus
input
depth
recovered scene radiance

52. Results: De-focus
input
depth
de-focus

53. Results: Video
output
input

54. Results: Video
output
input

55. Limitations Inherently white or grayish objects input our result
transmission

56. Limitations Haze imaging model is invalid – e.g. non-constant A input
our result

57. Summary Dark channel prior A natural phenomenon
Very simple but effective
Put a bad image to good use

58. Imatest
Imatest LLC is a company that produces image quality testing software and offers a range of consulting services. Imatest was founded by photographer/engineer Norman Koren in Boulder, Colorado in 2004 to develop software for testing digital camera image quality.

59. Wedge
The Wedge module (described in this document) can analyze any arrangement of horizontal or vertical (but not diagonal) wedges, using manual region selection.

60. ISO 12233:2000
The ISO 12233:2000 resolution test chart and its variants are by far the best-know charts with wedge patterns.

61. For the Wedge module, select the regions of interest (ROIs) to be analyzed.

62. ROI Fine adjustment window
The boundaries of the selected region should be inside the ends of the wedge (top and bottom, above) and outside the sides, allowing a little “breathing room”.

63. MTF and Aliasing onset results
MTF: Modulation Transfer Function
MTF and Aliasing onset results