Research And Implementation On Clearness Of Pathological Images

Pathological Image is an important basis for medical diagnosis,and treated as the gold standard for various diseases.However due to some external factors,the generated pathological image is typically existing noise and edge blurry,which directly affect the diagnosis.Aims at the problem above,we proposed a method to improve the quality of pathological image.The method can produce clearer pathological image for the pathologist,which has great research significance and practical value to the computer-aided diagnosis of pathological image.Consider the liver sinusoid is the feature of liver cancer,the research of liver sinusoid is also important of liver cancer.Based on studying and analyzing large amount of related domestic and foreign paper,our research object is H&E(Hernatoxylin and Eosin)staining pathological image.And this paper mainly researched the pathological image denoising and pathological image enhancement,besides,a special method was proposed for liver sinusoid according to its features.(1)On pathological image denoising,firstly,we introduced the location information to improve the similar block matching method;then we developed pathological image denoising model with isotropic filter and block matching;finally,we used alternating direction method of multipliers to solve the model.(2)On pathological image enhancement,we focus the measurement function of edge and texture.We replace the variance of pixel with variance of gradient,enlarged the range of variance and is convenient to select the threshold value.In addition,we introduced global gradient variance to make selecting of the threshold easier.(3)On clearness of liver sinusoid,firstly,we denoised the imge by orientation-selective filter,then,enhanced the image based on statistic feature,finally,we suppressed the fat droplets and other areas which are similar to the liver sinusoid based on morphological method.The experimental result about pathological image denoising and enhancement shows that,the pathological image denoised model we proposed in this thesis can denoise ideally.And the improved guided image filter can enhancement pathological image well.Meanwhile,the experiment of liver sinusoid clearness shows that our proposed mehods have been preferably improved.