On Study And Realization Of Median Filtering And Segmentation Algorithm Of Waterfront Images

Image denoising is an important research area of image processing.Noise will de-stroy the image information and make the follow-up work hard,so for the image polluted by noise should denoising before other image processing work. Image segmentation is another important research field of image processing,through the image segmentation, the image can be divided into target area and background area,do the follow-up work on-ly in target area can improve efficiency and reduce computation.The existing denoising and segmentation method is not very suitable for the waterfront image in this article,so based on the characteristics of waterfront image, we improve the existing denoising and segmentation method to make sure them can have very good effect for waterfront image in this article.The work of this article are as follows:1.According to the characteristics of impulse noise,we make some assumption-s for the waterfront image's pollution situation, on these assumptions,give a impulse noise detection methods,based on this methods we improve the existing color image filter method.Through experimental results,we can find the improved filter method can effectively filter impulsive noise of waterfront image in this article, so it meet actual problems's needs.2.Because of the waterfront image has a obviously waterfront boundary between target area (the whole water) and background, so can use this boundary to segment wa-terfront image.Based on the differences between waterfront boundary and other edge in large scale wavelet transform, we give a rapid waterfront boundary identification method and a waterfront image segmentation in this paper. Through experimental re-sults,we can find the image segmentation given by this paper can segment the whole water from the waterfront image rapidly and effectively.3.Usually,the waterfront image polluted by impulsive noise(pollution ratio not over 50%),so we add impulsive noise(density is 50%) to the experimental image for simula-tion actual situation.We use PSSMF method given by this paper to filtering the image polluted by noise in the first,then use segmentation method given by this paper to seg-ment image.Through the final segmentation results can found the method given by this paper can denoising and segment the waterfront image polluted by noise rapidly and effectively,that means the result in this paper can meet the need of the actual problem.This paper is divided into five chapters, specific content as follows:Chapter 1,introduce the background of the question studied in this paper,and intro-duce research situation of median filter and image segmentation.Chapter 2, introduce some basic knowledge in image processing, transform image into mathematical mod-el, and simple introduce wavelet theory used in this paper.Chapter 3, firstly introduce the impulsive noise model and existing color images median filtering method.Then,ac-cording to the characteristics of impulse noise in this paper, give a impulse noise detec-tion methods,based on this methods we improve the existing color image filter method .Last,give experimental results and a kind of image adaptive filter algorithm solves the construction methods.Chapter 4,give the image segmentation defintion introduce the existing basic method of image segmentation, then combines with the characteristics of the waterfront image need segment,explain why chose edge segmentation to seg-ment image.At last,, w give a rapid waterfront boundary identification method and a waterfront image segmentation based on the differences between waterfront boundary and other edge in large scale wavelet transform,also give experimental results. Chap-ter5,make a simple summary for the work been done in this paper.The innovation point in this paper is a rapid waterfront boundary identification method based on the differences between waterfront boundary and other edge in large scale wavelet transform.Compared with waterfront boundary identification based on fractal dimension,the method given by this paper can avoid iteration and heavy compu-tation.