Research And Implementation On Fog Image Restoration Technology

In autum and winter,fog is a common weather phenomena.Influenced by fog,the obtained imageis low contrast,not clear and can not reflect the true scence information.The out-door vision systembecome invalid under the influence of heavy fog,which may causes damages to some works.In this paper,the image dehazing algorithm based on the dark channel prior is studied.Thepre-estimated transmissiion map obtained by this algorithm exists many blocking effects,which can beeliminated by the soft matting algorithm.The soft matting algorithm needs the inversion of hugeLaplace matrix,which is computional complex and hard to implement.The algorithm,which combinesthe dark channel prior with the guided image filter,is proposed.The proposed algorithm replaces thesoft image matting algorithm by the guided image filter,which is the Jacobi iteration of the optimalsolutions for the soft image matting.Then the proposed algorithm is less computional complex andeasy to implement.For the further reduction of the computional complexity,a novel way to estimate thetransmissiion map is proposed,which obtain the exact transmissiion map and avoid the blockingeffects. he proposed method is effective and of less computational complexity. Meanwhile,research onthe median filter and kuwahara filter,which are the commonly used edge-corner preserving filter,isdone.The specified direction and multi-stage median filter is proposed,which overcomes the poorcorner preserving property of the median filter.By the redefinition of the outcome of the kuwaharafilter and gaussian-weighted approach to calculating the mean and variance of the sub-block,theproposed kuwahra filter has more effective result.Although the image image restoration algorithm has attractive dehazing result,the restored imageis not perfect which is overall dim,is of low contrast and the local details are not prominent.In thisrespect,Adaptive clipping histogram equalization is proposed.This algorithm can adapt the rate ofenhancement according to the noise of the local area.Then the restored image can be well enhacedavoiding the situation that some useful image information is drowned in the noise.The proposed algorithm is implemented on the personal computer and the Seed-Vpm6467,whichis a classic Davinci video processing system of TI company.