Transform Domain Rain Removal Methods using Dictionary Learning Approach: A Comparative Study

Rain removal from color images and videos is one of the challenging tasks in image processing. This paper proposes an efficient algorithm for the removal of rain from rainy images. This paper shows the comparison of different techniques that can be adopted to remove rain from color images. While several previous pieces of the research proposed different ways of obtaining a rain-free image by using a bilateral filter, guided filters along with the dictionary learning method, this paper proposes a different approach i.e., using hybrid l1-l0 decomposition for the separation of the image into low-frequency and highfrequency components. The non-rain image details are extracted from the high-frequency part based on the histogram of oriented gradient (HOG) features using dictionary learning. These non-rain details are then added to the low-frequency part to obtain the final rain-free image. This paper also shows the performance comparison of the proposed algorithm in the presence of different filters for obtaining low and high-frequency components.