1 Fuzzy modeling
(1)
The original R, G, B color of the pixel point P (original clear image)
The color of R, G, B observed by pixel point P (blurred image)
global atmospheric light
t(p) ϵ[0,1]，
The transmittance of the reflected light, determined by the distance d(p) from the point in the scene to the camera , is inversely proportional to the distance
: attenuation coefficient, determined by weather conditions, generally assumed to be 1 under typical fuzzy conditions ,
2. Static image dehazing
1. Estimate the atmospheric light A of the input image
2. Assume that the scene depth is the same in a block area of the image. Find an optimal transmittance t(p) for each patch area to maximize the contrast of the restored image
3. Minimize information loss due to pixel truncation while enhancing contrast.
4. Optimizing t by changing the block-based transmittance map to a pixel-based transmittance map using an edge-protected movable window filter to eliminate blocking effects.
5. From the input blurred image, consider the transmittance map and atmospheric light to restore the scene
2.1 Estimating atmospheric light (a hierarchical search method based on quadtree subdivision)
1. Divide the input image into four rectangular areas, and use the average value of pixels in each area minus its standard deviation as a score
2. Select the area with the highest score and subdivide it into four rectangular areas, repeat 1,2 . until the last selected area is smaller than our predetermined threshold
3. In the final selected area, 
the color vector with the smallest distance is the most atmospheric light
2.2 Estimating the optimal transmittance
1. Formula (1) is rewritten as follows:
When A is estimated to be only, it is only determined 
by t
2. Contrast calculation : Mean squared error (MSE) contrast
It can be seen from the formula that contrast is inversely proportional to t .
3. Loss function: Contrast loss function
4. t value
So have
illustrate
1: The essence of the t value optimization of the transmittance map in the back is not understood, so I will not upload it for the time being. I will update it later when I understand it.
2: There is still video stream defogging at the back of the article. I don't have time to watch it for now. I will update it later when I have time to understand it.
3: The original text Optimized contrast enhancement for real-time image and video dehazing, this article is translated from my own understanding