Defogging Method For Inland River CCTV Images And Videos

The CCTV monitoring system installed on the inland waterway, is a kind ofimportant means for the inland river maritime supervision at present. In the foggyenvironment, the images and videos captured by CCTV monitoring system, havesome degradation phenomenon such as lower contrast, color distortion and fuzzy,bring certain difficulties to monitor, and directly affect the play of the system as afunction of the eyes. At present, many scholars study deeply the images defogging,but due to the computation time of defogging algorithm is long, it is difficult tosuitable for application in the videos defogging. For the problem about low definitionto CCTV surveillance images and videos, the dissertation detailedly analyzes thecauses of the inland foggy degraded images and fuzzy mechanism, further studies thekey technology of inland foggy degraded images, the dissertation detailedly analyzesand improves the existing defog methods, and introduces new ideas to put forward aquick method for inland image to defog. The main work has the following fouraspects:1. Through analyzing the characteristics of inland foggy degraded images, andresearching the influence mechanism of the fog to inland foggy images, thedissertation sets up a kind of restoration model to be suitable for inland foggydegraded images. On the basis of this model, the dissertation clears the inland foggydegraded images.2. Put forward the k-means clustering method to estimate sky brightness value.This method firstly segments the sky area and non-sky area by k-means clustering,and automatically determines the clustering center of the sky, weights summation ofthe mean brightness value of the sky area and the brightness value of the most strongfog center, as ultimately the sky brightness value.3. Emphasisly research and analysis the defogging method based on Retinex,dark channel prior and depth information for inland image. And in view of the limitedapplication of these methods, the halo phenomenon and the defects of highcomplexity, the dissertation proposes a rapid defogging method based on HSI colorspace with the combination of filtering. On the one hand, on the basis of the fog without affecting the hue component of inland images, the dissertation recovers thebrightness component and the saturation component of inland images, respectively,makes the inland images after defogging retain more original information; on theother hand, Gaussian guided filter algorithm is used to optimize the airlight figures,greatly reduces the complexity of the algorithm, improves the operation rate.Experimental results show that the rapid defogging method based on HSI color spacewith the combination of filtering, can significantly improve the contrast and colordefinition of inland foggy degraded images, make the inland images after recoveryhave better visual effect, improve the visibility of CCTV monitoring system.4. In order to meet the real-time requirements of CCTV surveillance video, themotion vector estimation method is proposed to solve the transmission problem ofairlight figures between video frames, be able to achieve the speed of five frames persecond, realizes the videos to defog, and ensures the real-time performance during thevideo to defog. However, in that the dissertation does not consider video space andtime consistency, the videos after defogging will appear local dithering phenomenon.This is the place where the method need further study to improve.