Mechanism for Removal of Rain From Digital Images

The rain streaks in an image generally cause an imparity in the processing of the image in any image processing algorithm, this is a huge problem under a neural network that would consider these pixels as a part of the image, this might lead to error in the output. To overcome this imparity in images we use the L0 gradient smoothing method to remove the layer of the image, this layer corresponds to the rain pixels, which are typically low amplitude structures and hence removing these pixels by increasing the number of jumps in pixel format causing the overlap of unwanted pixels with much more important pixels and retaining the necessary sharpness of the image. This method also allows the removal of unwanted noise and increases the prediction rate with retention of the structures necessary for processing the image. This method is also useful in removing noise produced in the compression of animated pictures into JPEG format. The removal of rain pixels in any image could improve its processing quality and drastically improve the prediction probability. Finally, after removing these rain pixels, we have an image with low gradient and lower contrast, to improve the contrast and enhance the image we use HOG (Histogram Oriented Gradient System), this allows proper prediction and improves picture quality, this paper also aims at detecting persons in a frame after removal of noise from the input image, thus prediction of people in the given picture is much easier when the HOGS is implemented to it after L0 gradient smoothing is applied.