Research On Denoising Algorithms For Esophageal OCT Imges

Optical Coherence Tomography(OCT)technology has been applied in biomedicine and other fields.Since the combination of OCT technology and endoscopy,endoscopic OCT has been applied to imaging of esophagus and other tissues.The image is easily contaminated by speckle noise in the imaging process,which is a multiplicative noise related to the signal.This will interfere with the medical diagnosis and subsequent image processing of OCT images.Therefore,the effective removal of speckle noise in esophageal OCT images without loss of important details is of great significance.In this paper,algorithms related to image denoising are studied.Through the analysis of various algorithms in esophageal OCT image denoising,two of them are selected and improved.Firstly,aiming at the shortcomings of anisotropic diffusion algorithm,an anisotropic diffusion algorithm based on local variance(LVBAD)is proposed.The algorithm improves the diffusion function and diffusion threshold,so that the edge can be detected more accurately in the iteration process,and speckle noise can be removed better.Moreover,the algorithm also performs different filtering processes for pixels according to the classification of pixels:median filtering and improved anisotropic diffusion filtering.Secondly,according to the characteristics of speckle noise distribution of esophageal OCT images,Rice distribution is selected as the probability distribution function of the images,and an iterative Bayesian framework modified with bilateral filter(BFMBF)is proposed.The final filtering results can be obtained adaptively through the whole iterative filtering process.LVBAD and BFMBF algorithms proposed in this paper are both tested on synthetic images and OCT images of guinea pig esophagus,and they are compared with the related algorithms respectively.The experimental results of denoising of normal and pathological esophageal OCT images show that:the peak signal-to-noise ratio(PSNR)of esophageal OCT images was increased by more than 10 dB and the speckle noise suppression index was reduced to 1/2 of the original one.In addition,BFMBF algorithm also enhances the edges of normal esophageal OCT images,and the results show that the performance of speckle noise removal of BFMBF algorithm is better than that of LVBAD algorithm.Besides,BFMBF algorithm has also improves the accuracy and stability of image segmentation of esophageal OCT images,and makes automatic segmentation results and manual segmentation results more close.