Research On Selective Total Variation Filtering And Retinex Enhancement Methods For Medical Images

Because of the limitation of medical imaging system,medical image will inevitably appear noise and blur,so filtering and enhancing medical image becomes the important preprocessing step before medical analysis.The total variation(TV)model and retinex method have a wide range of application in image filtering and enhancement respectively.Based on these methods,this paper selects optical coherence tomography(OCT)images,ultrasound images,computed tomography(CT)images for filtering and enhancement experiments.In this paper,according to the different characteristics of the image,the corresponding selective filtering or enhancement method is proposed.The main research contents and innovations of this paper are as follows:(1)According to the characteristics of medical images such as retinal OCT images,this paper proposes a selective total variation filtering method which is based on fuzzy C-means(FCM)and shearlet transform to filter the noise in medical images.Firstly,this paper combines the total variation method with the shearlet transform method to protect texture details while filtering noise.Then,this paper introduces the FCM clustering method to divide the image into three parts accurately,then assign appropriate weights to the different parts of the image in order to achieve targeted denoising of images in different areas.Experiments in retinal OCT images and carotid ultrasound images show that compared with the simple spatial or frequency domain method,the proposed method can achieve good filtering effects on images in different regions while maintaining good edge preservation.(2)According to the characteristics of cardiovascular OCT images,this paper proposes a selective filtering method based on FCM,second-order single oriented partial differential equation(SOOPDE)and shearlet transform to filter the noise in medical images.After FCM clustering,this paper constructs the diffusion rate control function to modify the SOOPDE model.Then this paper uses the mask derived from FCM to extract the non-tissue region of the result which is from the modified SOOPDE,meanwhile it uses the mask derived from FCM to extract the tissue region of the shearlet transform result.Finally,with the iteration,the two parts are combined to obtain the final result.Experiments show that compared with the shearlet-TV method,the proposed method has a significant effect on the noise removal of both the non-tissue and tissue region of cardiovascular image,meanwhile the fine texture can also be preserved adequately.(3)To address the common problem that limited contrast and noise interference exist in medical images such as CT and OCT,this paper presents a selective retinex enhancement method based on FCM and shearlet transform to enhance the medical images.Firstly,this method performs shearlet transform on the organization region of interest which is from FCM,and then it uses the multi-scale retinex method to enhance the low frequency part,and uses the selective degradation diffusion method to filter out the existing noise in the image.Through experiments on 22 medical images,it can be proved that the proposed method can obtain enhanced images with higher contrast,less noise and more information than histogram equalization method and contrast limited adaptive histogram equalization method.