Research On Image De-noising Methods

Noise will inhibit people's comprehension to an image. The aim of de-noising is removing noise in the image, improving the degree of people's understanding to the image, and processing the image conveniently. The main task of this paper is introduction to the image de-noising methods. And farther study of some de-noising algorithm is given in this paper. At the same time, several new de-noising algorithms are proposed. The de-noising result of these new methods is perfect.Image noise and quality appraise-methods are introduced in the first chapter. De-noising methods of image in spatial domain are introduced in the second chapter. These methods include neighborhood-averaging method, spatial low-pass filter method, several images averaging method and median filter method. De-noising methods of image in transform domain are introduced in the third chapter. These methods include low-pass filtering algorithm in frequency domain based on Discrete Fourier Transform (DFT) and de-noising algorithm based on wavelet transform. Two new de-noising methods are proposed in the fourth chapter. One is a new filtering algorithm for removal impulse noise based on detection of noise points. It is a improved median filter method. Another is a new non-linear filter algorithm using DFT, which de-noises image through containing the coefficients of big module in the frequency domain. Ridgelet transform is introduced in the fifth chapter. It includes principle, implement method and application of this transform. At last, a new shift-invariant de-noising method using finite ridgelet transform is proposed in this article. Experiment shows that the new algorithm can remove image noise and preserve the image details well.Some usually used de-noising methods are introduced, and farther research of these algorithms is made in this paper. At the same time, some new de-noising algorithms are proposed. And experiment results of these methods are perfect. But what is needed to say is that noise used in the experiment is ideal, and it always adding Gaussian noise or pulse noise. But in fact, noise is more complex. So many de-noising methods should be used when we want to get ideal performance.