Defog Method And Quality Evaluation Of Image Based On Atmospheric Model And Bit-plane Decomposition

The application of computer vision and image processing has made substantial progress in all walks of life,which is playing an irreplaceable effect.However,the original part of the traditional approach has a method is too complex,inefficient,the results are not satisfactory,evaluation is too cumbersome and other issues.Therefore,this paper aims at this problem,in the theoretical principles and evaluation methods and other aspects of the method to improve,effectively improve the original defects.The main contents and new ideas of the thesis are as follows:(1)For the sky region which is difficult to deal with the defog problem,this thesis improved a new dark channel defog method.Dark channel algorithm is more effective,but for the use of matting technology makes the calculation of large amount,this thesis made a parameter improvement,instead of the original formula in the external minimization,making the calculation of reduced efficiency and efficiency;In this thesis,we use the idea of compensation to deal with the problem that this part of the pixel can not be dealt with the whole pixel at the same time in this algorithm,and the effect is improved by introducing the correction and the threshold.(2)In order to solve the defog problem in images contain the rich texture and abundant scene depth,this thesis proposed a method combined fractional differentiation with center-around function Retinex.Traditional Retinex algorithm has a long time consuming due to the large computational complexity,so the circular neighborhood is used instead of the square neighborhood,and the better processing effect is obtained.The method can deal with more pixels and because of the use of the center-around function,we use the small-area equalization process to make the computation decrease greatly.In addition,this thesis also use the fractional differential method to improve the image processing,for the rich texture and obvious image defog effect.(3)In order to solve the grey image defog problem,this thesis put forward a defog method based on bit-plane decomposition.In the aspect of traditional image processing method improvement,this thesis deals with the defects such as block and color distortion in the gray image processing under the traditional method.First,the image is processed by the high and low cap transformation calculation,and then the bit plane decomposition method to deal with the image,and then use Freeman chain code from the horizontal and vertical direction to strengthen the image boundary,for each bitmap histogram analysis,and according to the histogram results,select the most flat image as the final result,the actual results show that the quality of the image has been effectively improved?(4)For the non-reference defog image quality evaluation method,this thesis put forward a new method and conduct relevant experiment to test it.In the aspect of image objective quality evaluation,this thesis selects the feature image block of the experimental image,which uses the MVG model to quickly match and fit the original image,and calculates the distance of the relevant image with the distortion image,And the classical method based on the visible-side defog quality evaluation method are compared with this method.It is obvious that the experiments show that we have a more efficient method of quality assessment.