Research On The Key Techniques Of Adaptive Image And Video Dehazing

In outdoor environments,light reflected from object surfaces is commonly scattered due to the impurities of the aerosol,or the presence of atmospheric phenomena such as fog and haze.Photographs of hazy scenes typically suffer from the low-contrast and offer a limited visibility of the scene.It is an annoying problem for photographers since it degrades image quality.It is also a threat to the reliability of many outdoor vision applications under these poor weather conditions.Therefore,haze removal is highly imperative computer vision/graphics.In this thesis,we analyze the image degradation model and exploited the key techniques of haze removal in details,then our image and video dehazing methods are proposed.Our main research and contributions are summarized following:Traditional dark channel prior based haze removal schemes often suffer from the color distortion and generate halo artifacts in the distant views.To tackle these issues,we present an scene-adaptive single image dehazing approach via open dark channel model.First,we detect the image depth information and separate it into close view and distant view.Then,an open dark channel model is proposed to optimize the whole atmospheric veil,in which the values of close view are regularized by a minimum channel image while the distant parts are estimated by an appropriate lower constant.Accordingly,the transmission map of haze layer can be further optimized by guide filter and smoothed by domain transform filter.Finally,the haze degraded image can be well restored by the atmosphere scattering model.The extensive experiments have show that the proposed image dehazing approach has significantly increased the perceptual visibility of the scene and achieved a better color fidelity visually.We present an efficient single image dehazing approach via multi-scale wavelet analysis and open dark channel model.First,we have heuristically found a generic regularity in nature images that the haze is typically distributed in low frequency spectrum.Benefited from multi-scale wavelet decomposition,we use an open darkchannel model to remove the haze effect in low frequency part.Then,by considering the coefficient relationships,we utilize the estimated transmission to further enhance the texture details in high frequency parts adaptively.Finally,the haze-free image can be well restored via the wavelet reconstruction of the recovered low frequency part and enhanced high frequency parts correlatively.The extensive experiments have shown its outstanding performance.A real-time video dehazing algorithm,which preserve the temporal and spatial coherence and reduces visual artifacts,is proposed in this work.Assuming that a scene point yields highly correlated transmission values between adjacent image frames,we develop the temporal coherence term.Then,we add the temporal coherence term to the incremental transmission term to define the overall cost function.By tradeoff the overall cost function,we obtain the optimal transmission.To reduce the possible visual artifacts in the image recovery process,we future propose jointly recovering the final image by using the relatively noise-free hazy image to reduce noise in the dehazed image.Experimental results demonstrate that the proposed algorithm is sufficiently fast for real-time applications and effectively removes haze and yields high quality output videos.