A Research On Image De-noising Methods

Image is an important source of information, it may help people to understand the connotation of information through the image processing. The image noise will hinder the understanding of an image. The purpose of de-noising is to remove the image noise and improve the awareness level of the image, so that we can make the further processing to the image. The digital image de-noising involves optical system, microelectronic technology, computer science, mathematical analysis and so on. It’s a very comprehensive interdisciplinary science, now its application is very widespread. In the medicine, the military, art, the agriculture and many other domains, it has a very extensive and wide application. Matlab is one kind of highly effective engineering calculation language, in value computation, data processing, imagery processing, neural network, wavelet analysis, it has the widespread application. The main work of this thesis is to give the introduction of image de-noising methods, make a further study to the de-noising algorithm and conduct simulation experiments to analyze their de-noising.The first chapter describes the overview of the development of digital image de-noising algorithm and research summary of some of the traditional de-noising algorithms, introduces the latest research in the field of the de-noising algorithm. The second chapter gives a brief introduction for image noise and image quality assessment methods. Most image noise is a basic salt and pepper noise and gaussian noise, in the existing de-noising algorithm mostly noise are them. This article not only describes and analyses the two mathematical models of the noise and the gray value histogram, but also gives the de-noising quality criteria. Image quality evaluation methods include two ways:one is subjective evaluation, and the other is objective evaluation. The general objective evaluation method is widely used. The third chapter introduces the spatial domain image de-noising methods and transforms domain de-noising methods. And include neighborhood average method, airspace low-pass filtering method, the average of multiple images, median filtering method, the adaptive median filtering and wiener filtering method. This paper introduces the principle of these algorithms, the filtering process, carries out a detailed analysis and discussion of the characteristics of the median filter and adaptive median filter and wiener filter, the filtering process, as well as its shortcomings and advantages, and by Matlab we carry out a detailed simulation of the above algorithm and analyze the results of the de-noising performance.The fourth chapter gives a detailed study of the latest image analysis tool:the wavelet transform. It introduces the concept of wavelet function, wavelet reconstruction and wavelet transform. Wavelet transform is divided into discrete transform and continuous transform. It also introduces the concept of nature and the decomposition process of the wavelet packet. The article focuses on the image de-noising algorithm which bases on wavelet transform and wavelet packet transform and analyses the characteristics of coefficient distribution of the images after the two transform and the selection of the threshold function principle. The fifth chapter gives detailed simulation analysis of the de-noising performance in the de-noising algorithm which bases on wavelet transform decomposition level, bases on two-dimensional wavelet packet transform, bases on two-dimensional stationary wavelet transform decomposition level de-noising and bases on the median filter different types of de-noising, and compares their pros and cons of de-noising effect. The sixth chapter is the conclusions and outlook. This part gives my own views of the development trend for wavelet de-noising.The full article summarizes and introduces some of the frequently used method of image de-noising which bases on the image space domain and transform domain, the basic principles of wavelet threshold de-noising and wavelet packet de-noising are carefully studied. I analyses and simulates several de-noising methods. Finally, through theoretical analysis and experimental results, the article discusses a variety of factors affecting the de-noising performance of a complete de-noising algorithm. In the actual image processing, it gives data reference and a basis for selection and improvement of wavelet threshold de-noising method and median filtering.