Restoration Of Image Degraded By Dynamics Weather

Dynamic weather image restoration research is full of challenge. Rain, snow andother frequent inclement weather seriously degrade the performance of video image,which greatly reducing the value of the image, causing a serious impact on theperformance of the visual system. For target recognition and detection, the impact ofrain and snow cannot be ignored. Thus, the image affected by rain must be recovered,minimizing the impact of rain in image application, ensure that the system can workwell.First study the physical property of raindrops and spatial distribution etc. Due tothe uniform distribution characteristics of raindrops, try to use several kinds of classicfiltering algorithm for rainy days recovery. Based on the Summarizes of the advantagesand disadvantages of existing algorithms, select two different methods for furtherresearch：One is frame-difference method with good real-time and the other isclustering algorithm which removal the rain well.In the study of frame differential method, analyze the raindrops characteristics indifferent color space. For only the V-channel of image interference by raindrops inHSV color space, only need to processing the raindrops in V channel can achieve theeffect of the three-channel simultaneous processing which greatly improving theoperating speed. Therefore, proposed an improved frame-difference method for rainimage restoration based on HSV color space. The existing three-frame difference andfive frame difference method has missing inspection problem sometimes. To solve theproblem introduce the difference between the mean value and mid-value as a constraint,the experimental results indicate that the algorithm has Good effect for static rainimage restoration. For the Fuzzy C-Means algorithm, since it required scanning pixel of each framepoint by point in the image, which reducing the processing speeds. To solve thisproblem, we apply Image parallel processing technology to enable computers’multi-core participate in parallel processing parallel, indeed improve the operatingspeed. In order to enhance the effect on the Fuzzy C-Means algorithm for large rain,first improved raindrop removal method. Then according to the uniform distributioncharacteristics of raindrops, studied the filtering techniques. It shows that the medianfilter have certainly help, so combined the fuzzy C-means algorithm with medianfiltering better completed the image restoration in case of large rain.