关键词: 图像恢复, 脉冲噪声, 小波变换, 大数定律, 中心极限定理

Abstract: The rightness of recognizing impulse noise is the key to the effectiveness of denoising.Wavelet transform can detect impulse noise in signals based on the principle that it can detect signal’s singular point.Experimental study shows that the total noise identification accuracy is affected by the error rate that a pixel polluted by impulse noise may be identified as a normal pixel when wavelet transform is used to identify impulse noise.In order to effectively address this issue，a statistical theory based method called MIVP method is proposed for identifying impulse noise in digital image which can make up for the weakness that a noised pixel may be identified as a normal pixel by wavelet transform.An image impulse noise filter is designed based on the wavelet transform combined with MIVP method.Such a filter can effectively improve the effectiveness of filtering impulse noise without increasing its time complexity.

Key words: image restoration, impulse noise, Wavelet Transform（WT）, law of large numbers, central limit theorem