Purpose of the invention: in order to provide an improvement on uneven foggy images, the invention provides a defogging method combining LSD secondary segmentation and deep learning. – This is an improvement of uneven foggy images. The invention provides a method for defogging using LSD secondary segmentation and deep learning. This includes acquiring a foggy image data set, secondary segmentation, thin fog processing, dense fog processing, and the final defogging result.
LSD secondary segmentation divides the image into uniform blocks and updates the threshold for segmentation to ensure uniformity.
Thin fog processing and dense fog processing are applied to an U-Net model to restore clear images. – The method proposed by the invention for blocking the unevenness of foggy images improves the performance of defogging. It uses U-Net, an improved median filter, and total variation smoothing. It improves on the existing DMPHN method of blocking the foggy image evenly. The invention can also solve halo and blocky artifacts and retain image details. Compared to the existing method, it has better performance and more visual appeal.
The method can be used in a variety of scenarios including civil, industrial, and defense applications.
