Research On Elimination And Atmospheric Haze Influence Of Image

Computer vision technology has a widely applied in all areas of our daily life.However,visual acquisition equipment will easy failure with the drop qualitative bad weather(fog,haze,rain,snow,etc.),which make the image mass and contrastreduce,color offset and distortion,etc..As a result, it has covered or blurred of the features in the image,great influence on subsequent processing of the image.Image haze-revomal as a key part of image process,mainly used in outdoor video surveillance, military topography and exploration, unmanned automated driving,etc..Therefore,to research the algorithm of haze image restoration,will reduce the influence of the bad weather,has a vital practical significance for the engineering application and scientific research.Single haze image restoration has many methods,mainly divided into image enhancement algorithm which is based on non-physical model and image restoration algorithm based on physical model of the atmosphere.Image enhancement does not consider degeneration causes,instead of specific needs of certain information to enhance image and diminish or eliminate the unwanted image information.But image restoration analyses from the perspective of physical causes of the atmospheric dispersion model,which can achieve recovery and high degree of reduction.Traditional of those two methods will lead to poor visual effect,the fault brightness dimming,edge details,making it difficult to practical application.For the above,in this paper has presented a L-2 Norm haze-removal algorithm,which has based on the variational framework of Retinex theory,Dark Channel Prior(DCP) and Atmospheric Veil Prior,enable recovery image with good visual experience.The constraint will lead to restored image mist based on L-2 Norm,unable to effectively identify the boundary between adjacent objects,bring some difficulties to further target identification work,also presents a L-0 Norm of single image haze-removal algorithm.The main contents are as follows:1. Further investigation of atmospheric imaging model,make the common light intensity which is constant in simplify model applied in non-constant. Systematic summarization image restoration based on atmosphere in recent years,analysis andusing the typical theory of a prior knowledge,establishment a novel single imagehaze-removalobjective functions based on L-2 Norm.2. He’s dark channel prior knowledge will make the restorate image dim,and operation with lots floating-point.In this paper,we presented a L-2 Norm haze-removal algorithm,which has based on the variational framework of Retinex theory,Dark Channel Prior(DCP) and Atmospheric Veil Prior,enable recovery image with good visual experience.Finally,using the classic-least-squares optimization method, iterative solution of the unknown component and get the values of inverse recovery eventually clear images. The simulation results shown that this algorithm has a good restoration.3. For image recovery based on L-2 norm constraint solver easy to make the scene after restoration information edge details vague, is not conducive to the separation and extraction problems edge structure information object, this paper edges of objects based on sparse prior knowledge, via L-0 norm constraint the restory image,to replace the image of the object restored L-2 norm constraint, in order to maintain the edge structure information to recovery image, and the corresponding threshold value iteration algorithm. Simulation results show that the proposed algorithm dehazing have a better results to recovery image.