Degradation model and group sparse representation-based foggy day image restoration method

Abstract

The invention discloses a degradation model and group sparse representation-based foggy day image restoration method. According to the degradation model and group sparse representation-based foggy day image restoration method, on the basis of research on a foggy day atmospheric scattering model, the depth variation law of each pixel and the law of the light change of the pixels caused atmospheric light scattering are analyzed and summarized, so that a foggy day image degradation operator is designed, and a foggy day degradation model is constructed; on the basis of the degradation model, a group sparse representation method is adopted to perform training, so that group dictionaries corresponding to each group are obtained; an SBI (split Bregman iteration) method is adopted to solve a sparse coefficient; and finally, a restored image is expressed by the group dictionaries and the sparse coefficient. According to the degradation model and group sparse representation-based foggy day image restoration method of the invention, the foggy day degradation model and the group sparse representation method are combined to calculate an image restoration result; the local sparseness and non-local self-similarity of the image are fully utilized; and therefore, it can be ensured that the restored foggy day image has high contrast and clarity definition.

Description

technical field [0001] The invention relates to a foggy image restoration method based on a degradation model and group sparseness, and belongs to the technical field of foggy image processing. Background technique [0002] Severe weather such as rain and fog will have a great impact on many industries. For example, in the transportation industry, whether it is expressway or aviation, rain and fog will increase the risk of traffic accidents. The purpose of image defogging is to remove the influence of weather factors on image quality to restore a clear image, which is a type of image restoration. This topic has been a research hotspot in the field of image processing and computer vision in recent years. [0003] The essence of many image restoration problems is to design effective regularization terms to accurately describe the prior knowledge of the image. In recent years, with the development of sparse representation theory, the sparse representation model has become a pr...

AI Technical Summary

Problems solved by technology

This kind of method requires more prior knowledge and lacks universality; (2) Another part of the defogging method based on non-physical model often has complex restoration process and difficult parameter setting, and the restored image often appears The color is oversaturated, and the Halo effect will appear in the place where the color of the scene changes suddenly

Images