Research On Several Improved Median Filter Algorithms

This paper focuses on a difficult problem of the image processing: Image Denoising. In the process of image transmission, there will be many external or internal various reasons which can make noises on image. Image Denoising algorithms can effectively reduce the noise and enhance the visual effect. Meanwhile, it can also improve image quality and make the image more realistic. As a difficult problem in image processing, image denoising still has significant meanings.This paper briefly introduces the development of image denoising and research of present situation first. Then, advantages and disadvantages of the existing image denoising algorithms are analyzed in this paper. Based on these theories, this paper finds the main difficulty of image denoising, which is called high density noise and image details preserving. Based on salt&pepper noise, it proposes two main contributions, as follows:First, it proposes new energy models based on energy minimization and multiple median filter, which is named energy model, such as model El and model E2, and then, proposes a new multiple median filter. According to solve the local neighborhood energy minimum, two models judge whether pixel grayscale value is substituted. The new multiple median filter can not only change the single way of original neighborhood mean filter algorithm, but also reduce the possibility of the good pixels be mistaken for the noise pixels. Meanwhile, it can change the threshold free. This filter is particularly effective in suppressing Salt&pepper noise. The other applications of this filter are that it can make up for global energy function removing the binary image noise and image segmentation, etc.Second, this paper proposes a new noise detection algorithm. And then, this paper proposes three new filter algorithms, such as Sliding filter window Standard Median filter(SSM),Sliding filter window Center Weighted Median filter(SCWM) and Sliding filter window Tri-State Median filter(STSM). These third filters tend to preserve more image details at the expense of suppressing noise, even high density noise.