An Alogrithm Of Image De-Noising Based On Improved Tail Mean And Wavelet Transform

As the information carrier for our daily communication,an image can objectively reflect objects to help us in knowing,understanding and changing the world.An accurate and highresolution image passes on information efficiently and assists us to recognize internal laws of objects and their correlations.However,it is a pity that,an image may be inevitably affected by noises in its acquisition,reception and preservation,causing damages to its original data and even seriously affecting our work or study.To obtain the desired information,it is very necessary to consider the image denoising as the key part of image data processing.Generally speaking,an image noise is not just of one singly type,instead,it is a combination of Gauss noise and impulse noise,there exist different type noise in different location.While,the denoising method in pre-treatment actually aims at either Gauss noise or impulse noise,so it often gives unsatisfactory results.This paper focuses on the improvement of trimmed mean and wavelet function used in image denoising.For images contaminated Gauss noises and impulse noises,using the improved trimmed mean and improved wavelet function to remove the above two kinds of noises.The specific work is summarized as follows:Firstly,the classification of image noise and common noise models are briefly summarized.We introduce the traditional image denoising methods,including: spatial domain filtering,frequency domain filtering and optimal linear filtering.The theory of wavelet threshold function is expounded.Secondly,we focuse on improving tail-cut mean,wavelet transform and ridgelet transform.The main principle of Tail-cutting mean is to sort the data,remove the maximum and minimum values in the data,and then get the mean of the remaining data.But we know that when the information of an image is comparatively large,the gray value of the image will vary greatly,and then the image will be the most.The maximum or minimum value may be image information.At this time,if the Tail-cutting mean is used,the image information will be lost.On this basis,an improved Tail-cutting mean method is proposed.This method mainly judges the number of maximum and minimum values,and then judges the number of maximum or minimum values according to the given conditions.At the same time,it is applied to suspicious noise points.Secondary test can protect the original information of the image while filtering the noise better.As to Gauss noise,The traditional threshold method is analyzed,we improve a new threshold function at the same time,the performance of the function is analyzed.Gauss noise is filtered by combining wavelet transform and ridgelet transform.The experimental results show that the improved tail-cut mean method can remove salt and pepper noise well,and it also has a good effect on Gauss noise.The improved threshold function combined with wavelet transform and ridgelet transform can filter Gauss noise well.Finally,this paper introduces the new method that is to detect image noises under the mixed noise influences by using the way proposed herein.For the selected sub image,the noise type is identified based on its features,and remove the mixed noises correspondingly by the aforesaid two methods.At last,this paper demonstrates the effectiveness of this new method by experiments.