The Research Of Binarization Method Based On Non-uniform Illumination Images

There is information which is interested in and not interested in by people in one image.Inorder to highlight the target of interest and remove unuseless information,binarizaiton method isusually used to extract the features of images for identification to make images to be black andwhite binary ones.Global binarization methods are usually used for effectively binary segment in the imageswhose illumination is uniform and have a greater difference between target andbackground.However, the images taken by people are weak target image which have a samllerdifference between target and background due to the actual scene illumination conditions orenvironmental restrictions, so it is invalid for these images to use the traditional globalbinarization methods for segmentationMany scholars at home and abroad have already proposed many solutions for thenon-uniform illumination image binarization.the main methods are processing the image byblock, processing illumination of image background previously, using the classic localbinarization methods and so on. The effect of many experiments have indicated that thetraditional method of uniform illumination of the the weak target image processing is notacceptable, there are many disadvantages such as:the separation of target and background isincomplete or the loss of a valid target,low processing efficiency and so on.This paper has summarized the current binarization methods,and a method based on thesingle-column processing and the improved OSTU is proposed to solve the binarization problemof a class of non-uniform illumination,According to the illumination character of the images, thescheme of single-column binarization is firstly presented,the initial binarization threshold valuesof each column of the images are determined by the classical OSTU method.Then a way isdesigned to adjust the column of the binarization threshold adaptively based on each columnillumination intensity,and show the analytical formula to calculate the binarization thresholdadjustment amount of each column.After that,in the view of parameters of the formula need to bedetermined artificially,an improved OTSU binarization method based on template matchingalgorithm is proposed to calculate the parameters in the formula automatically,so as to arrive theadaptive threshold adjustment amount of each column in the class of non-uniform illuminationimages to optimize the initial binarization threshold.The method has been proved to be effectiveby the image processing results of a group of defective billet images,and show the processingresults compared with the effect of other classic binarization methods.