目的 为了克服彩色图像去噪后存在的特征模糊，研究基于双边滤波的自适应彩色噪声图像去噪方法。方法 使用二维离散小波变换（DWT）对含噪声的彩图图像进行近似分量、水平细节分量、垂直细节分量和对角细节分量等4个方向的分解。根据DWT各方向分量归一化后的方差比例，利用RBF神经网络构造双边滤波系数模型确定不同方向的最佳去噪系数，提出彩色噪声图像自适应去噪方法（DWT-ABF），并将该方法与常规方法作对比。结果 在不同噪声类型以及混合噪声失真情况下文中方法都能有效地去除噪声，并同时保留图像细节信息，且与其他方法相比，文中方法去噪后的图像都具有更高的PSNR值。结论 文中方法克服了传统双边滤波无法自行确定最佳参数的缺陷，同时也良好地解决了去噪图像特征模糊的问题。

The work aims to study the adaptive color noise image denoising method based on bilateral filtering for the purpose of overcoming the feature blurring in color image after denoising. The color noise image was decomposed into approximate component, horizontal detail component, vertical detail component and diagonal detail component by the two-dimensional discrete wavelet transform (DWT). According to the normalized variance ratio of components of DWT in each direction, the RBF neural network was used to construct the bilateral filter coefficient model to determine the best denoising coefficient in different directions, propose an adaptive denoising method for color noise image (DWT-ABF), and compare this method with the conventional method. In the different types of noises and mixed noise distortion, the proposed method could effectively remove the noise and preserve the detail information of the image. Compared with other methods, the images denoised based on DWT-ABF had higher PSNR value. The DWT-ABF overcomes the defect that traditional bilateral filtering is unable to determine the optimal parameter, and it also well solves the feature blurring of denoised image.