Researches Of The Image Dehazing In Real Time

Image dehazing has been a hot topic in computer vision and digital image processing field in recent years. There are two main methods:The one based on the atmospheric scattering model and the one does not base on physical model. In this paper, we have an in-depth perspective on defogging method based on atmospheric physics model. The main contents include the following aspects:Firstly, we elaborate the status and significance of the image dehazing technology gave a classification comparison of defogging algorithms, illustrate the advantages and disadvantages of each algorithm. According to the characteristics of atmospheric scattering model, in this paper we study the methods of image dehazing based on atmospheric scattering model. Through in-depth and detailed studying of the formation of haze and principle, we gave a general understanding of the reasons why image is degraded in poor weather conditions, such as haze and fog. The dehazing theory is documented with image characteristics.Secondly, we studied the two main defogging methods based on the model of atmospheric scattering. Based on maximize local contrast dehazing method and single image haze removal using dark channel prior. Through in-depth study of the principles of these two algorithms, and expounded their advantages and disadvantages respectively.Maximize local contrast dehazing method bases on a truth that outdoor image taken in clear weather has a larger contrast than the one with fog. We model the airlight in the Markov random field and compute the airlight using graph cut algorithm. The results of using maximize local contrast dehazing method have a larger contrast than others, the color is more saturated and have halo effect on depth discontinuity.An error occurs when the transmission rate is less than the local highlight objects against dark priori. This paper studies the defogging method based on dark priori and region segmentation. This method can effectively correct the error of transmittance using original method. Experimental results show that this algorithm can effectively refine the original transmittance get corrected result. The defogged images both in the resulting edge detail and in terms of color are better than using the original method.Finally, we studied a real-time image defogging method based on physical model of atmosphere and bilateral filter. The first step is according to the principle of atmospheric physics model with dark priori method and using bilateral filter to get the transmit map. After solving the transmit map directly, the transmit map can be take into the model to calculate the defogged result, without refining the transmit map. The algorithms increase the computational efficiency of the algorithm. Experimental results show that this method satisfies the real-time requirements of speed. At the end of this paper, methods of evaluating defogged image’s quality are given in order to assess the image quality after dehazing objectively.In this paper we use Matlab and C++ to implement all the algorithms.