FAST EFFICIENT ALGORITHM FOR ENHANCEMENT OF LOW LIGHTING VIDEO: Based on De-Fogging Algorithm

This paper presents a FAST EFFICIENT ALGORITHM FOR ENHANCEMENT OF LOW LIGHTING VIDEO based on defogging algorithm.

The authors of the paper observed that the inverted image (inverted image) has similar properties to the fogged image, such as:

The above two properties are unique to fogged images.

Therefore, we can use a mature de-fogging algorithm for low-luminance image enhancement.

Specifically, the approach is as follows:

Where, is the brightness of the atmosphere, is the brightness of the image acquired by the camera, is the brightness of the original image or scene.

Based on [1] , we can get:

where is the scattering coefficient of the atmosphere, is the depth of field of the pixel.

Where is set to 0.8 in the algorithm, and is a small region centered at 9 in the algorithm.

To obtain the brightness of the atmosphere, the authors chose the 100 pixels that are the largest in the smallest value of the RGB channel of the image, and then chose the pixel with the largest sum of the RGB values of these pixels to be the most estimated value.

Note that the areas we need to enhance are objects in the foreground, such as houses, cars, and so on, and we need to avoid over-enhancing the background areas, such as the sky.

So, here we need to adaptively adjust the content of the image to focus on enhancing the foreground content. Therefore, an intermediate variable is introduced here:

Then, the picture to be recovered can be calculated by the following formula:

[] The paper also introduces how to accelerate the video, which is ignored as it is not the focus of the research in this blog, and interested parties can check the original article.