Research On Low Illumination Image Enhancement Technology Based On Dark Channel And Wavelet Fusion Algorithm

At present,the demand for image acquisition is everywhere in people’s life.However,the image acquisition equipment is greatly affected by the light,and the images collected in the poor lighting environment show dim,uneven illumination,information loss and other problems,which seriously affect the visual effect of the images and lead to the image cannot be identified and used.Therefore,the study of low illumination image enhancement algorithm has a very important application value for regional security,traffic command,large crude oil storage tank and other scenes.The main research contents and achievements of this thesis are as follows:(1)After the simulation analysis of a variety of classical low illumination image enhancement algorithms,aiming at the problems of poor universality,low saturation,detail loss,image whitening and so on,an improved Retinex and CLAHE fusion algorithm is proposed.When extracting illuminance images,this algorithm combines the advantages of the 2-order Butterworth low-pass filter,which has smooth characteristic curve and fast attenuation rate at the cut-off frequency,to replace the traditional low-order Gaussian filter with it.Meanwhile,the guided filter is used to replace the high-order Gaussian filter to alleviate the problem of detail loss,and the medium-scale Gaussian filter is combined to form a hybrid filter.After the image is transferred into HSV space,the improved Retinex algorithm filters the enhanced illuminance image obtained by the V channel,and pre-enhances the reflected image by using the improved threshold CLAHE algorithm.The two enhanced images are weighted and fused to form the enhanced V channel image.The adaptive stretching method is used to alleviate the image whitening problem for the S channel.Finally,the enhanced HSV image is transferred back to RGB space to obtain the final image.Through multi-group experimental simulation and comprehensive qualitative and quantitative analysis,it is verified that the improved Retinex and CLAHE fusion algorithm significantly improves the visual perception,and significantly improves the average gradient,information entropy,contrast and other indicators.(2)In view of the limitations of the above improved fusion algorithm in terms of peak signal-to-noise ratio and structural similarity,inspired by the similarity between foggy image and pseudo-fog image after inversion of low illumination image,this thesis further proposes a low illumination image enhancement algorithm based on dark channel de-fogging and wavelet transform fusion.Aiming at the two major problems of halo artifact and color distortion in the traditional dark channel after de-fogging,the algorithm proposed to use the multi-scale small window dark channel edge preserving method to refine the selection of transmittance in the area of image shading and depth of field to alleviate the halo artifact problem,and to optimize the selection of atmospheric light value to reduce the influence of over-white pixels in the image such as reflection spot and light source.Ease the color distortion problem;Aiming at the problem that the improved dark channel enhancement algorithm has too much noise,the wavelet transform threshold denoising method combined with the improved threshold function is proposed to de-noise the initial enhanced image.The method has small error,strong detail retention ability,and effectively improves the peak signal-tonoise ratio of the image.Through several experimental simulations,it is verified that this algorithm has better visual perception than other algorithms,and its comprehensive performance in five indexes is higher than other algorithms.(3)In order to verify the feasibility of the application of the proposed algorithm in practical engineering,this thesis relies on the cleaning monitoring system project of large crude oil storage tank to carry out enhancement experiments on low-illumination images inside the tank.The results show that the proposed algorithm can greatly improve image quality,improve image contrast and highlight image details,no matter for the image with uneven illumination or the image with extremely low illumination.So that low illumination image can be fully and effectively used.