Application Of Image Defogging Based On YCbCr Color Space

With the area of people expand and the pollution of the sky increases,it becomes difficult to get a clear picture in bad weather.So it is very important to get the images when the extreme weather happens.In recent years,the images restoration technology of digital multimedia technology has become the current hot issues.This paper is based on the atmospheric scattering model.And put a lot of enery on the reseans of image quality.In the last proposed two different principle of fogging method.These two methods have achieved good results.The main work of this paper is as follows:1)The images defogging method based on improved dark channel prior was proposed.This method starts from the physical model.Firstly,the dark channels prior of two different color spaces are obtained by the prior rule of dark channels;Secondly,the Quadtree method was used to estimate the atmospheric light value for two different color Spaces;Thirdly,calculating the mean transmittance of RGB color space and Y channel transmittance of YCbCr color space;Then,using the bilateral filtering to refine the transmission image.This method can improve unclear edges;Finally,two images come from different color space are given different weights to duplicate the original image.Experiments show that this method can effectively solve the problems of high detail loss,insufficient image saturation and unclear edge,and can produce more real and clear images compared with the existing image defogging methods.2)An image defogging method based on multi-scale convolutional neural network is optimized.Using large synthetic data Reside sets as the original data sets for our foggy day images.Image defogging is realized by learning the mapping relationship between fog map and clear image of RGB color channel and YCbCr color channel through neural network structure of multiple convolution.Firstly,multi-scale convolution is used to check the haze characteristics of the image for extraction.Secondly,two residual modules are connected to prevent gradient explosion during training.Finally,the clear image is restored by the deconvolution module.Through a large number of data,it is shown that this method has strong identification,high contrast and good saturation effect in both indoor and outdoor natural scenes.In summary,a large number of experiments on the images of the test set show that the two image defogging methods proposed in this paper are superior to other methods.