- This project is mainly based on the work of Kaiming He who is my favorite researcher.
- I do suggest you use python 3 which is both stable and widely used nowadays.
- The code also works well with python 2 since I just use numpy and matplotlib.
- The requirement to the version of these packages is not strict, because we just use the basic function.
- The tools used to read image is matplotlib which is in the order of RGB while cv2 is in the order of BGR.
- I want to remind you that even the same operator can have different meanings between matrix and array in numpy.
This method was mainly put forward by Kaiming He, and then it was published at CVPR 2009 which was rated as the best paper.
This method help us get a fine haze removal image in a fast way with less cost compared with matting method.
From paper, we can see that the matting laplacian is the inspiration of guided image filter, they are both ingenious and attractive.
- I don't want to package the code so that you can see the details, and you are welcomed to give me suggestion.
- The variables used in my code are as same as these in official papers so that you can get it easily.
- The function
haze_removal()contains each step used in image de-hazing, I wrote this just for convenience.
-
Tiananmen Square : the original image comes from Kaiming He's paper
-
The Forbidden City : the original image comes from Kaiming He's paper
-
New York : the original image comes from Kaiming He's paper
-
HongKong : the original image comes from Kaiming He's paper
-
Airport : the original image is downloaded from internet
-
The Street : the original image is downloaded from internet
-
A380 is taking off : the original image is downloaded from internet
