Research On The Methods Of Restoring Degraded Image Due To The Snow Weather

Machine vision systems are becoming increasing widely used in the field of industrial control, intelligent transportation, aerospace technology. Rain, snow and other inclement dynamic factor reduced the quality of the image sequence and performance of the machine vision system, which bring huge problem about video surveillance, target recognition. Therefore, it has important and broad application to remove the raindrops and snow for outdoor systems for improving the quality and visual effects of image, this paper presents the research work under snow in the video.At first, studying snowflake visual perspective characteristics, spatial and temporal distribution characteristics based on the physical characteristics of the snowflake shape, size, speed, we try to use filtering method to recovery a single image. By implementing existing basic algorithms to analyze the advantages and disadvantages, and we study frame difference which has a wide range of target detection basis for further research.In the study of frame difference, we analyze the basic principle of the frame difference method for snow video, using the frame difference of two, three and five to detect and process respectively. According to their own physical imaging properties of snow, adding color constraint processing improvements for preliminary testing, after removing snowflakes of snow video using the frame difference method respectively. Finally, comparing the treatment effect and the running time respectively, it laid the foundation for the subsequent removal algorithm.Traditional color dynamic weather degraded image restoration algorithm is done in the RGB color space, which requires three components information for processing respectively. Based on the analysis of snow on the color attributes, we transform video image from RGB color space into YCb Cr color space, founding snowflakes mainly consisting in Y component. Using this feature, we propose to improve the five-frame difference method based on the color space, using the area information and the orientation angle statistical information to detect snowflakes for improving. Finally, in order to remove snow area of snowstorm video, we use median filter for the final step. Experimental results show that our algorithm has better restoration effect.