Research On Images Enhancement Algorithms Under The Foggy And Hazy Weather Condition

Many computer applications for outdoor surveillance system, such as road surveillance, security monitoring, etc., are required to accurately extract the image features. However, due to the atmospheric particles scatter, image captured in foggy weather is often badly degraded, which suffer not only serious visual quality, but also difficulties with feature extraction. Consequently, outdoor monitoring systems cannot run normally, bringing serious security problems. Therefore, it has important practical significance to sharpen hazy images by an effective and fast defog method.This paper first introduces the atmospheric scattering model, analyzes the cause of image degradation, and studies in depth of the image de-haze algorithm based on the dark channel prior. In order to solve the problem of high computational complexity of soft matting, this paper proposes an improved joint bilateral filter to optimize the transmission. The complexity of our algorithm is only a linear function of the input image pixels numbers; this allows our method applicable for real-time requirement.The result image obtained from the image de-haze algorithm based on the dark channel prior has problems of low brightness, low contrast and indistinct low details. In order to solve this problem, this paper proposed hue-preserving adaptive clip histogram equalization based on the analysis of the relationship between the transmissions of de-haze image and noise amplification extent. Experimental results show the image obtained by our method has better visual effects and restrain the noise at the same time.As for errors existing in the edges of transmission map which has variable field depth, this paper proposed a new haze removal method from single image based bilateral filter. We first make full use of the edge-preserving of bilateral filter to get the accurate atmospheric veil. Then, in order to solve the problem of distortion of the bright areas, we proposed a weakening defogging strength method. Finally, the haze-free image is recovered by inverting the atmosphere attenuation model. Experimental results show that our algorithm can get good defogging effect, especially the distant scene and places where depth changes abruptly.This paper studies the de-haze algorithm based on Retinex algorithm in depth. In order to improve the speed of Retinex algorithm and visual effect of the result image, this paper propose novel MSRCR algorithm based on Gaussian recursive filter and a linear stretching method to enhance the contrast of the result image. The experimental results show the proposed algorithm can achieve better defogging effects and higher speed of operation. This paper classifies surveillance images by introducing the idea of visibility. Making full use of the advantages of the proposed de-haze algorithm, the proposed algorithm implements adaptive haze removal and is verified its validity and practicability by experiments.