Study On Image Clearness Technique In Bad Weather

The outdoor bad weather will affect the normal work of the visual system, making the quality of image and video captured in this situation degrade greatly. One of the most frequent weather is the haze, which is an important factor affecting the quality of image. In haze weather, atmospheric scattering phenomenon will occur, this often blurs the captured image and reduces identification of information. The application value of haze-image is greatly reduced, which will interfere with the normal work of others areas related to image processing. Moreover, in recent years, the haze and fog weather are found frequently in our country. Dehazing technology of image has become a more and more important research topic, which has remarkably practical significance.Due to the dehazing algorithm is involved in the randomness and complexity of haze weather, this increases the difficulty of the research. Although there are a lot of dehazing algorithms in the home and abroad, they all have a number of limitations to some extent. These algorithms can be divided into enhancement method based on non physical model and restoration method based on physical model in essence.Firstly, in this paper, we present a haze removal method based on physical model. Firstly, we calculate the environmental light using the improved White Patch Retinex algorithm. Secondly, we estimate the initial transmission by minimizing the information loss. Finally, according to the haze image degradation model, the clear de-haze images are obtained. Experimental results show that the finally dehazing images obtained by proposed algorithm are bright enough and no additional operation for enhancing brightness is needed. Moreover, there will not appear serious problems of color distortion in the sky areas, and the method also has good applicability and robustness.Secondly, we convert the video sequence’s RGB color space into YUV color space, and only process the luminance(Y) component, without modifying the chrominance(U,V) components. This greatly improves the computational speed, thus suitable for processing video sequence, and the processing speed up to 20-22 frames per second for haze video with good real-time performance.Lastly, this paper also studies the defogging method of surveillance video. Because of the invariant features of the background of the video surveillance, we can separate the background and foreground and process them with the dehazing method. Then fusion the clear image to a video surveillance. Because of the background image is static and unchanging, so only dehaze the background image for a one time, this greatly reduces the computation time of video defogging algorithm, and has very good practical significance.