|
| 基于Retinex和双边滤波的NSCT域遥感图像去雾 |
| Remote Sensing Image Defogging in NSCT Domaim based on Retinex and Bilateral Filtering |
| 投稿时间：2024-02-27 修订日期：2024-04-08 |
| DOI： |
| 中文关键词: 图像去雾 非下采样轮廓波变换 双边滤波 梯度增强 Retinex算法 |
| 英文关键词:image dehazing non-subsampled contourlet transform bilateral filtering gradient enhancement retinex algorithm |
| 基金项目:福建省中青年教师教育科研项目 |
|
| 摘要点击次数: 40 |
| 全文下载次数: 0 |
| 中文摘要: |
| 针对遥感图像易受天气影响整体对比度差、轮廓细节不清、视觉效果不佳的现象，本文基于非下采样轮廓波变换（Non-subsampled Contourlet Transform,NSCT）算法，并结合Retinex算法与双边滤波实现图像去雾增强。首先，对有雾图像的各个通道分别进行NSCT分解获得各通道的高频与低频子带；然后，对低频子带采用Retinex算法和线性增强处理，提升图像灰度的动态范围，提高亮度均匀性，而高频分量先通过双边滤波在保留边缘的同时滤除噪声分量，再通过梯度增强处理，增强图像边缘轮廓；最后，重新融合处理后的低频子带系数和高频子带系数，实现遥感图像的去雾增强，并通过白平衡算法保持图像色彩一致性。将实验结果通过主观和客观方式进行有效性评价，结果显示本文所提算法的去雾图像中均值适中、平均梯度提高较多，因此本文算法在去噪的过程中可以很好的保留高频细节、改善对比度，图像质量明显提升。 |
| 英文摘要: |
| This study proposes an enhancement method based no Non-subsampled Contourlet Transform (NSCT) combined with Retinex algorithm and bilateral filtering algorithm to solve the shortcomings of lack of brightness, blurred edge details, and poor visual effect of remote sensing images that are prone to be affected by fog. Firstly, each channel of the fog image is decomposed by NSCT to obtain the high frequency and low frequency sub-bands of the image. Then, the Retinex algorithm and linear enhancement are applied to the low-frequency sub-band to enhance the dynamic range of image grayscale and improve brightness uniformity. The high-frequency sub-bands are first filtered by bilateral filtering to retain the edges while filtering the noise, and then enhanced by gradient enhancement to enhance high-frequency details such as edges. Finally, the processed low frequency and high frequency sub-bands are refused to obtain the defogging enhanced image. The experimental results are evaluated by subjective and objective methods. The results show that the average value of the defogging image by the proposed algorithm is moderate and the average gradient is greatly improved. Therefore, the algorithm can effectively retain image details, enhance image contrast, and achieve better visual effect while removing noise. |
| 下载PDF阅读器 |
| 关闭 |