Calculation of precipitation intensity using processing the sequence of video frames

Abstract

The article describes the method for automatic precipitation detection and its intensity estimation from a sequence of video frames. The convolution operation used for filtration is optimized in order to decrease computational complexity. The accuracy of precipitation intensity estimation is improved by accounting for cross-correlation between sequential video frames. The proposed method has reduced computational complexity. The results of the study are demonstrated on a few images of different size. The computational complexity of proposed algorithm is appropriate for real-time implementation.