Image de-Fog algorithm

Preface:--first to explain the background, has been doing FPGA hardware, because the business needs to do image processing. Started by a small white, from Reading Gonzalez began the image processing road. In the middle also experienced a lot of bumpy twists and turns, not much to say the open picture.

Read some of the image to fog papers and materials, found that most of the "Dr. He Keming (Hong Kong University) proposed by the Dark channel transcendental theory to do research", I also wrote about Dr. Ho's article, really convinced of the feeling.

The paper is named "Single Image haze removal using dark channel prior"

In this paper, the parameters are explained as follows: I (x) is the fog image (the image to be fog), J (x) is a fog-free image that needs to be restored, a is the global atmospheric light composition, t (x) is the transmittance. Now the known condition is I (x), in order to be able to seek the target value J (x). The author puts forward the theory of dark channel transcendental-that is, the dark channel of normal non-fog natural landscape image is tending to all black, which should be the value of 0 per pixel.

The steps of the algorithm are:

1 Find the input fog image (each channel of the image participates in the operation, RGB) of the dark channel data.

2 Take A, (I chose the maximum in the Dark Channel as a)

3 drawn by 1 and 2 T (x)

4 fog-free diagram from 3

Here are a few comparison charts.

Write here today, follow up to perfect.

Image de-Fog algorithm