基于非下采样轮廓波变换和加权引导滤波的遥感图像增强

Remote Sensing Image Enhancement Based on Non-Subsampled Contourlet Transform and Weighted Guided Filtering

针对部分遥感图像整体亮度偏暗、边缘细节特征模糊和可视性不够理想的缺点,提出了一种基于非下采样轮廓波变换(NSCT)与加权引导滤波的增强方法来改善图像质量.先利用NSCT获取图像多尺度子带图像,再对低频子带图像采取全局映射调整亮度,利用加权引导滤波器代替Retinex中的高斯滤波器获取细节分量和基础分量,同时采用比例因子调整两分量在低频子带图像中的比例;采用改进的自适应贝叶斯阈值和非线性增益函数增强各个高频子带图像;最后将各子带信息通过NSCT逆重构得到增强图像.与传统图像增强算法相比,该方法在清晰度和信息熵等方面有所提高,较好地保留细节特征,明显提高视觉效果.

This study proposes an enhancement method based on non-subsampled contourlet transform(NSCT)and multi-scale guided filtering to solve the shortcomings of lack of brightness,blurry edge details,and unsatisfactory visual effects for partial remote sensing images.First,the multi-scale sub-band image was obtained using NSCT.Then,global dynamic mapping was applied to a low-frequency sub-band image to adjust the brightness.Accordingly,a weighted guided filter was used to replace the Gaussian filter in Retinex to obtain the detail and base components.Scale factor was utilized to adjust the ratio of the two components in the low-frequency sub-band image.The adaptive Bayesian threshold based on the features of each direction and the enhanced nonlinear gain function were employed to improve the high frequency sub-band coefficients.Finally,the processed sub-band was inversely reconstructed by NSCT to obtain an enhanced image.Compared with traditional enhancement algorithms,the proposed method herein improves definition and information entropy,preserves detail features,and enhances the visual effect.