Research On Image Sequence Haze Removal

With the rapid updating of science and technology, computers and multimedia technology have been widely used. The importance of security has been pained more and more attentions. Surveillance systems as one of the most effective means of keeping security has been widely used in the field of traffic safety, bank and so on. However, the captured images will be with low contrast and less color information when the weather condition is bad, such as fog, haze or smoke. The surveillance system cannot work well. It may cause significant loss to these fields. Haze removal is really highly desired in these situations. The research in this issue focuses on haze removal of image sequences and does following researches:This paper researches the model of the formation of a haze image and studies the strengths and weakness of the single image haze removal method based on dark channel prior.A single image dehazing method based on dark channel prior and wavelet transform is employed. This method is based on the observation that the haze mainly has an effect on the low frequency component of a haze image. Consequently, single image dehazing can be approximated and simplified by removing the haze in the low-frequency component of haze image, which helps reduce the runtime cost.Median filter and Non-local Means filter are introduced to refine the transmission map. The experiments confirm that both of them have superiority than soft matting algorithm.Based on Just Noticeable Difference(JND), Human Vision System(HSV) is relatively sensitive to the change of luminance at the medium of the background luminance and less sensitive at the dark or bright areas. Usually, the dehazed image looks dim. So the dark image enhancement methods based on quadratic and parabola function are introduced.Kalman filter is adopted to estimate the atmospheric light of the image sequences. There will be flicker between the image sequences dehazed by the single image dehazing method. Based on the observation that the atmospheric light is usually invariant in a period of time, Kalman filter is employed.By combining the experiment results and the theory, a conclusion is given. Thehaze removal method proposed in this work is effective and can meet the needs of a real-time monitoring and control system to a certain extent. It can be a solid base for the further investigations on image dehazing. Finally, the future work is discussed.