Aiming at the problem of image dehazing in more bright and dark areas

there are problems that restored image being darkness and missing image edge detail information

an image dehazing algorithm based on Dehaze Net and edge detection mean-guided filtering is proposed. Dehaze Net can be used to train a convolutional neural network

and to estimate the initial transmittance map of the haze image based on four haze-related features.The initial atmospheric light value map is obtained by local atmospheric light value

instead of a single atmospheric light value

to avoid the darkened image being restored. An edge detection mean-guided filtering algorithm is proposed to optimize the initial transmittance map and the initial atmospheric light value map

and retain the edge detail information. Finally

the dehazing image is restored based on the atmospheric scattering model. The experimental results show that the subjective visual effect of the dehazing image is better

and the objective evaluation results of image information entropy

peak signal-to-noise ratio and structural similarity are better than other comparison algorithms. The algorithm can solve the above problems of restoring image darkness and losing image edge detail information

making the dehazing image clear and natural.