基于小波的遥感图像薄云去除的研究与实现

【作者】 王恒进；

【导师】 赵忠明；

【作者基本信息】 西北工业大学 ， 计算机应用， 2002， 硕士

【摘要】 遥感图像的薄云和雾霭覆盖是遥感图像处理中最常遇到的一种情况，有效的去除云覆盖的影响是遥感图像预处理阶段的重要任务。本文在这里所介绍的去除云覆盖现象的方法属于数字图像盲恢复的研究领域。 本文结合中巴资源卫星地面应用系统图像应用处理分系统“课题的需要，采用数字图像变换的方法来去除无辅助信息的遥感图像薄云覆盖，通过对薄云覆盖的遥感图像进行变换域处理，进行高通滤波去除云覆盖所在的低频信息，通过同态滤波并对图像进行空频分析来增强图像背景。然后在此基础上，通过引入小波变换，利用小波变换的多分辨率分析来进一步改善效果。 小波分析是目前国际上公认的最新时间-频率分析工具，由于其“自适应性”和“数学显微镜性质”而成为许多学科共同关注的焦点，本文利用小波改善传统数字图像处理方法，取得了一定效果。 本文提出多种评价模式，包括目视判别和统计分析，经过大量实验证明，本文所探讨的方法对于去除无辅助信息的大面积含云遥感图像的去云处理具有很好的可操作性和有效性。 本文主要的创新之处在于： 1．基于传统遥感图像去云的思想，建立了一种新的图像薄云去除数学模型，并在此基础上提出了空频分析通过灰度补偿改善去云效果，同时结合IHS变换改善彩色图像薄云去除，取得了很好的结果。 2．将传统高通滤波和同态滤波的思想与小波变换相结合，充分利用小波变换的多分辨率分析，更好的保持了图像的细节信息。更多还原

【Abstract】 Cloud cover noise often occurs in remote sense image processing. It is a very important task to remove the effects of cloud cover. How to realize it is the necessary stage of bind restoration of remote sense image processing.This paper introduces a way to remove the effects of cloud by frequency transform. Here, information of cloud cover is removed by high-pass filtering and information of background is enhanced by homomorphic filtering and spatial-frequency analysis. Furthermore, a new way is particularly discussed based on wavelet transform.After strict tests, we find that multiresolution analysis based on wavelet transform can improve the effect of the method of Fourier transform. The results show this approach is effective when there is massive cloud cover on the remote sense image.Wavelet analysis is internationally recognized up to the minute tools for analyzing time-frequency. It is chiefly due to the "adaptive feature" and "mathematical micro-telescope feature". This paper introduce the application of wavelet to image processing.This paper has such special points as:1. Based on the idea of removing cloud cover, a new mathematical model has been constructed. This paper introduces a way to improve the effect of removing cloud cover by spatial-frequency analysis and a new method to improve the effect of removing cloud cover on color image by IHS transform.2. Wavelet transform and conventional high-pass filtering method have been combined. Using multiresolution analysis of wavelet transform, details of the result image after being removed cloud cover are greatly improved.更多还原