Researching On De-noising Algorithms In DR Images

With the rapid development of IT and computer technology, modern medicalimaging technology assisting medical diagnosis also has made a great progress. Digitalradiography has been widely used in clinical medicine for its advantages of small doseof radiation, high sensibility, high resolution images and fast processing. However, thenoise produced by the system itself and external surroundings during the process notonly brings down the quality of images, but also interferes the medical diagnosis.Hence, how to explore the more appropriate de-noising method for DR images hasbecome the emphasis and difficulty in this research. The characteristics of noise in DRimages are learned preliminary by means of analyzing the composition and principle ofDR imaging system. And the related researches have been carried on in connectionwith various noises. In this paper, the main research tasks and achievements are asfollows:1. Study the basic theoretical knowledge of DR imaging system. The noiseproduced by DR imaging system is understood by the analysis of imaging proceed ofDR. Combined with the research of the main components of DR imaging systemflat-panel detector, the object of study in this article—impulse noise and Gaussiannoise is determined.2. Study the median filtering method based on adaptive center weighted. Firstly,apply scalar quantization method to deal with image space to avoid repeated filteringto the same image, which improves the detection of noise points and de-noisingefficiency. Secondly, adopting the algorithm of adaptive weight setting saves thesetting threshold can not only achieve effective removal of noise points, but also makethe detail information in the image fully reserve. And then, combined with vector spacepartition method, iterative filtering is used for every space area to ensure the bestresults of removing the noise points. Finally, a series of experiments have verified thatmeanwhile, the algorithm can well suppress impulse noise while preserving the detailinformation in the original image, and possessing strong stability.3. Study the anisotropic diffusion filtering method based on adaptive estimation ofthreshold. Due to Gaussian noise characteristics in DR image, a kind of anisotropicdiffusion filtering method, based on adaptive estimation of threshold, has been put forward. Firstly, through the analysis of P-M model, the diffusion coefficient isimproved to optimization, which slows down the convergence rate of the diffusioncoefficient, avoids the defects of incompletely filtering the noise caused by the rapidconvergence and improving the efficiency of filtering of the noise. And then, set thethreshold function as one dimensional function by using adaptive threshold estimationmethod to make sure that it is dynamically adjusted according to the pixel of imagethreshold. Finally, a series of experiments have verified that meanwhile, the algorithmcan well suppress Gaussian noise, while preserving the detail information in theoriginal image, and achieving strong stability.