An Optimized Image Dehazing Algorithm Based on Local Entropy Theory

Aiming at the problems about color distortion and obvious hierarchical changes of the sky regions in the results of current image dehazing algorithms, this paper proposes an optimized method. We understand the process of image dehazing from the perspective of information gain, and uses the local entropy and brightness features of the image as the main quantitative characterization to recognize the sky area. According to daily life, the sky area in images has the characteristics of small local entropy and high brightness. This conclusion is verified by the analysis of multi-sample images. After obtaining the approximate sky area recognition results, the open operation and the mean filter are used to optimize the sky image. Finally, the image optimized is used as the guide image, and the fusion is performed between the sky image and dehazing image obtained by the traditional method. Compared with the existing methods, the method proposed by us has the following advantages: strong recognition basis and high accuracy; theoretically improved speed; robust. This paper verifies the effectiveness of the method by subjective comparison and objective analysis of four samples using different methods.