Research Of Image Enhancement Technology On Non-uniform Illumination Of The Mine

With the development of industrial automation, efficient and safe production of coalmining receives more and more attention, and coal-mining industry video monitoring systemhas been widely used in the major collieries in land and abroad. Images collected with miningcamera will appear different lights as the location changes, which will lead difficulty in imageanalysis and decision-making after image information acquisition, so before image analysisand decision-making, image information enhancement should firstly be in need.The dissertation aims in mine complex environment at further research intocommonimage enhancement theory and technology which can improve the non-uniformillumination image visual effect, analyzes the existing short comings in practical applicationand presents a method of image enhancement. The thesis mainly includes the followingcontent.First for the characteristics of non-uniform illumination of coal mine images, enhance theairspace of pixels. By comparing the direct contrast enhancement algorithm, the sharpeningmask enhancement algorithmand histogram equalization enhancement algorithms, theanalysis results can be obtained. Image enhancement method of histogram equalization inprogress of low illumination image has good visual effect, but as in the processing ofequalization will be rounded gray levels, making some grayscale loss, easyly to cause imagedistortion.Then put forward the image two-dimensional wavelet decomposition in frequencydomain, and respectively make enhancement to high frequency and low frequencydecomposed images. Make threshold processing to low frequency coefficient, and the highfrequency part adopts the fusion rule based on partial energy matching degree to enhance thedetails. Experimental results show that this method can do very good in interest region detailsof users, but the smoothness of reconstructed image is very low. Finally Retinex theory and common several implementation methods are introduced, andit analyzes thecommon method limitations in progressing of the non-uniform illuminationimage. Innovatively it puts on the basis of the Retinex separation of luminance image forwardthat,first makeimage filter with Gaussian kernel function, adjust the primary filtering resultsby changing the scale parameter, and then continue brightness image separation to the primaryfiltering results by iteration vector method.The experiment results prove that the method caneffectively increase the image contrast, reduce background noise, and improve the overallvisual effect of image.