Gradient-Based Photorealistic Rendering of Rain Streaks for Image Processing Applications

To record criminal activities who loot company properties in various places such as parking lots, outside the work place, banks, industries etc., surveillance systems are installed. Under aggressive and bad weather conditions, the images recorded contains rain streaks and are difficult to conceptualize. In these situations, images will help the police in investigation. Removal of rain droplets from still images has been an effective research topic for various image processing applications. In this paper, to the best of our knowledge, we use Histogram Oriented Gradient (HOG) to extract rain droplets from an image. After extraction, K-means clustering and SVM classifiers removes the common rain patterns present in the image. Finally, the performance is evaluated by calculating the PSNR value for original image and comparing the values with the PSNR of rain removed image, which are more or less found to be equal.