Affiliation:
1. Ryazan State Radio Engineering University named after V.F. Utkin
Abstract
The relevance of the task of detecting and eliminating rain and snow particles in television images taken with static cameras is shown. The procedure for detecting rain and snow particles on video images, which is supposed to be used as part of an algorithm for reducing the visibility of precipitation, has been developed. The approach developed is presented in the form of three-stage classification of frame pixels into zones with moving objects and areas of stationary background, distorted and undistorted by precipitation particles. The procedure proposed has been studied and optimal values of selected accumulated frames for its correct operation have been determined: 100 frames for video images with rain; and 140 frames for video with snow. The gain of the approach developed in comparison with known by levels of errors of the first and second kind is 0.8 % and 3.4...9.1 %. The described procedure can be used in TV systems where the delay of 3...5 seconds relative to video source is not critical.
Publisher
Keldysh Institute of Applied Mathematics