Research Of Algorithm Of Haze Image Sharpening

Outdoor captured images usually have weak quality because of the atmospheric turbidity media (such as molecules, water droplets, etc.). Fog, haze, smoke can cause such phenomena by absorption or scattering. In addition, the resulting light is also mixed with atmospheric light (light reflected by air molecules surrounding environment), contrast and color fidelity of haze image are declined. The degradation of image is varied with space, because the extent of atmospheric scattering is related to the distance between scenery and camera. According fog degradation model, image recovery from a single foggy image is ill-posed problem. But now the clarity of outdoor vision systems has become more important, outdoor subsequent image recognition processing need a clear picture as the foundation. Moreover, due to environmental pollution and other reasons, the fog and haze emerge more and more frequently in our country. Therefore, the haze image sharpening is a hot and challenging problem.In this paper, we deeply analyse the image degradation model, and study the problem of fog image clarity based on the Dark Channel Prior and Retinex theory. We propose the changing scale Retinex algorithm for haze removal based on depth map and non-uniform stretching Retinex algorithm, that largely eliminates fog Retinex enhancements to some drawbacks such as slow speed and weak adaptability to different images. This paper gives a lot of ideas to the study of the haze image restoration, especially the idea combines the Retinex enhancement and image depth together for the first time. The main content of the paper is as following:First, we deeply analyse the basic theory of fog image clarity, which is haze imaging model. The model can be divided into two parts:the reflected light attenuating optical imaging model and atmospheric model. We detailed analysis of these two models to prepare for the follow-up work.Second, we introduce some effective defog method such as the Dark Channel Prior, Retinex enhancement and Fattal’s defogging algorithm. After comparing the results of different method, we analyze and summarize their advantages and disadvantages and provide a theoretical basis for the follow-up work.Third, we explain the principles and procedures of the changing scale Retinex algorithm for haze removal based on depth map and non-uniform stretching Retinex algorithm. Prove that the proposed algorithm is better by comparing the results of differenet methods. Pixel entropy and contrast is proposed as objective evaluation criteria to evaluate the processing results. Also, compare the complexity of several algorithms.Finally, a summary of the work is carried out, we points out the shortcomings of the paper and see the inadequacies as a key work in the future.