Drizzle Detection through Histogram of Radon Projection (HRP)

Abstract

In hilly regions, frequent drizzling interrupts the daily activities of human beings and can also be deceptive. In this work, we develop a feature descriptor by Histogram of Radon Projection (HRP) to detect drizzle/rain. To compute this feature descriptor, initially we detect fine edges in an image utilizing the sobel gradient operator. Then the resultant image is divided into smaller cells and for each cell we estimate the count of radon transform values for different orientations and the weighted average for each transform coefficients are accumulated into bins. Finally, we use this radon feature descriptor to detect whether drizzling occurs in a given frame or not using Support Vector Machine (SVM).