Research On Denoising Algorithm For Images With High Density Salt And Pepper Noise Based On Median Filter

Images are objective visual reflections of nature things, and also important carriersof access and dissemination of information. With the rapid development of computerscience, information science, statistics and other disciplines, digital image processingtechnology as one cross discipline is developing very fast and has integrated into peoplelives in various fields. Noise and original signal are always intertwined, which impactsvisual effect of image seriously. Therefore, the study of image filtering algorithms hasalways been a hot research of image processing field. Median filter as a simple andpractical nonlinear filtering method is widely applied to the research of removal of saltand pepper noise. To improve the filter capability of filter under the condition of highdensity of salt and pepper noise and protect image details, is the contradiction that isdifficult to balance, and it is the important and difficult point in the process of imagedenoising. For that purpose, the following works were done in this paper:(1) According to the characteristics of salt and pepper noise distribution, aneffective noise detection strategy based on extreme values was deeply studied in thispaper. Firstly, the image pixels are divided into signal points and points that might bepolluted by salt and pepper noise, which is basing on extreme values. And thendistinguish image detail and noise points in pixels that might be polluted by comparingto the average value in the neighborhood. Then, filtering is done at those pixels thatwere detected as noisy. The noise detection strategy is not only improving the efficiencyof filtering algorithm, but also reduces the omission and false detecting rate of noisesignal greatly.(2) A filtering measure based on rules extracted from local pixel information wasproposed in this paper. Multi-rules were extracted to recover noise according to thecorrelation between pixels around noise signal and the characters of filtering window.It’s a good way to protect image details and reduce error.(3) By introducing an improved filtering method based on adaptive variablewindow, the performance of the proposed filtering algorithm based on local pixelinformation extraction rules was further improved. In view of the shortcoming of thetraditional filter algorithm for the highly polluted images, this paper put forward anadaptive selection filter window size control strategy according to noise density, andonly the signal points were sorted in the window. Thus, it not only improves the filteringcapacity, but also reduces the spread of noise signal during the filtering process again.(4) A kind of simulation platform based on OpenCV is designed in this paper. Thesimulation was built by using the advantages of OpenCV and reconfigure Visual Studio2008. On the OpenCV simulation platform, few denoising algorithms for images werestudied. Experimental results show that: the proposed method has a better filteringperformance under high density of salt and pepper noise, can protect image detail andimprove peak signal to noise ratio better.