Abstract: Digital images are often contaminated by noise during image acquisition and transmission.Sometimes，noise is a mixed one of Gaussian noise and impulse noise.Unfortunately，existing denoising algorithms are often designed for removing single Gaussian noise or impulse noise.In this paper，an efficient method for removing Gaussian-Impulse mixed noise is proposed，which can be described as follows：Firstly，Boundary Discriminate Noise Detection（BDND） is applied to detect pixels corrupted by impulse noise，then the noisy image is filtered according to a decision map and an image corrupted by Gaussian noise is obtained；Secondly，Bayes Least Squares-Gaussian Scale Mixtures（BLS-GSM） is utilized to denoise the obtained image.Experimental results show that the proposed method can efficiently remove mixed noise.

Key words: Boundary Discriminate Noise Detection（BDND）, Bayes Least Squares-Gaussian Scale Mixtures（BLS-GSM）, mixed noise

摘要： 数字图像在获取和传输过程中常常会受到噪声的污染，有时会同时受到高斯噪声和脉冲噪声的污染。然而现有的去噪算法大多针对单一的高斯噪声或脉冲噪声，在处理混合噪声时无法取得令人满意的去噪效果。给出了一种去除数字图像中高斯-脉冲混合噪声的有效方法，去噪过程分为两个步骤：首先采用一种称为边界判定噪声检测的脉冲噪声检测方法检测出混合噪声中的脉冲噪声，对噪声图像作中值滤波后得到一幅受高斯噪声污染的过渡图像。然后用贝叶斯最小平方-高斯尺度混合模型对过渡图像进行滤波得到降噪后的图像。实验表明，同现有的其他去噪方法相比，该方法能够更有效地去除混合噪声。

关键词: 边界判定噪声检测, 贝叶斯最小平方-高斯尺度混合模型, 混合噪声