将引导滤波与提升小波相结合提出了一种多尺度引导滤波方法，以实现在平滑图像细节的同时保持图像边缘不模糊。该方法通过提升小波法对将图像进行多尺度分解，即将信号分解成一个低频子带和多个高频子带。在提升小波重构过程中，利用引导滤波平滑每个尺度的低频信息并保持其边缘不模糊。最后，针对滤波后残余的细节，对提升小波重构后的平滑图像再次进行引导滤波，以便进一步平滑图像细节。将多尺度引导滤波应用于暗通道去雾先验理论并进行了主、客观评价。结果显示: 多尺度引导滤波能够深层次平滑图像细节，保持边缘完整性，从整体上提高了图像的对比对和视觉效果，有效恢复了场景信息并保留场景的边缘信息。另外，该方法改善了客观评价指标，其对比度增强系数指标平均提升了0.1以上，场景结构相似度平均提升了1以上， 而LOE(Lightness Order Error)参数降低了10以上，满足了去雾应用的视觉需求。

With combining guided filter and lifting wavelet effectively, a multi-scale guided filter was proposed to smooth the details of a color image and to keep the edges of the image unambiguous. The lifting wavelet was used to decompose the color image in multi-scale. It means that the image was decomposed into a low frequency subband and a plurality of high frequency subbands. In the reconstruction process of the lifting wavelet, the low-frequency information of each scale was smoothed by the guide filter and the edges were not blurred. Finally, the reconstructed image was processed by the guided filter once more to remove the residual details as soon as possible. The proposed multi-scale guided filter method was applied to the image haze removal using dark channel prior and its processing results were evaluated in subjectivity and objectivity. The results show that the multi-scale guided filter smoothes the image's details and maintains the edge integrity. It improves the visual effect of the image in whole, enhances the image contrast and visual effect. Moreover, it recovers the scene information and preserves the edge information of the scene. At the same time, the objective evaluation indexes are improved, in which, the contrast enhancement coefficient index is improved by 0.1 or more, the restoration ability of the scene structure information is increased by more than 1 and the Light Order error(LOE) is reduced by more than 10. It satisfies the visual needs of the application in image dehazing.