Research On Color Image Dehazing

With the development of technology, computer vision and image processing have been widely used in many aspects, such as intelligent video surveillance and security systems. As we all know, our current climate is getting worse. Some cities are seriously affected by the haze of the weather, resulting that the images or videos captured by the outdoor computer monitoring system have many quality problems, such as low contrast, fuzzy information and color distortion. These problems have adverse influence on subsequent image processing, such as target recognition, feature extraction, etc. Therefore, it is very important to study the color image dehazing.At present, research on the algorithm of color image dehazing has obtained some achievements. There are two main kinds of algorithms. The first kind is based on image enhancement. The second kind is based on image restoration.Image enhancement algorithms use some methods to highlight the contrast and details of the image. Its advantage is flexibility. In order to achieve accurate image dehazing, the corresponding parameters are calculated according to the atmospheric scattering model in image restoration algorithms. As the exact parameters should be calculated, the complexity of this algorithm is high.Therefore, how to achieve an equalization in terms of the effect of restoration images and the real-time performance of the algorithm for image dehazing is a problem ,which needs further study in this field.In this paper, we focus on the research of color image dehazing and solve the problem of image clarity. Firstly, the current research status of this field at home and abroad is summarized, and the classical dehazing algorithms are introduced. Then, the key theoretical basis and technical points of this paper are introduced, such as atmospheric scattering model, foggy image degradation model, dark color prior and estimation of atmospheric scattering model parameters. Quality evaluation of restored images is introduced too. Finally,based on these theoretical basis and key technology points, two new image dehazing algorithms are proposed. The proposed algorithms effectively realize the clarity of the foggy color image. In addition, the performances of the proposed algorithms are compared with other dehazing algorithms comprehensively and fairly.The innovation of this paper mainly includes the following two aspects:·A new dehazing method based on global least squares and image enhancement is proposed. The algorithm introduces the global least squares filter into the image dehazing and combines the image restoration with the image enhancement. This algorithm makes the recovered fog-free image get the natural color while enhancing the contrast of the image. Compared with other algorithms, the results show that the image restored by this algorithm can fully recover the details and color in the scene. Both added visible edge ratio and the normalized gradient mean of the visible edges show that the contrast of the image restored by this algorithm is obviously improved.·A new dehazing method based on atmospheric light map and adaptive manifolds is proposed. Firstly, the whole image is divided into three parts according to the sky area of the image, and the corresponding atmospheric light can be obtained for each part. Then, a gradient atmospheric map is constructed by linear interpolation. Then the filter based on adaptive manifolds is used to optimize the transmittance map. Finally, the fog-free images are reconstructed based on the inverse process of the atmospheric scattering model. This algorithm uses accurate methods to obtain the atmospheric light and transmittance map. Therefore, the brightness and clarity of restored image can be wholly improved. In addition, both added visible edge ratio and the normalized gradient mean of the visible edges show that the proposed algorithm has better dehazing performance.