Detection and removal of rain and snow from videos based on frame difference method

摘要

Dynamic bad weather such as rain and snow will reduce the qualities of the image and video, and will affect on the performance of the visual system seriously. So dynamic weather image restoration technology is a challenging research content. According to the optical physical characteristics of raindrops/snowflakes, we studied the classic video frame difference algorithm detects rain/snow, and on its basis, identifying initial detection snow binary connected region, adding the constraints of area and direction angle to further distinguish raindrops/snowflakes region. At last, we use three algorithms named two-frame difference, three-frames difference and five-frames difference to remove raindrops and snowflakes, and compare the removal effect and the running time.