基于地物光谱矢量空间的遥感图像去噪方法研究

发布时间：2019-05-18 10:07

【摘要】：ETM+遥感器对地观测获取的遥感图像，存在大气干扰高频分量、前期图像处理留下的残余误差及来源不明的其它误差，在实际应用中需要去噪处理。为了提高图像清晰度，学者们探索了多种方法，其中低通滤波效果较好，被遥感图像处理软件广泛采用，但滤波效果仍不理想，地物内部弱显清晰，地物边沿弱显糊化，弱化噪声同时弱化了信号。去噪方法尚待改进，新方法尚待探索，希望新方法能弱化噪声、增强信号、输出精度较高较清晰的地物图像。本文结合低通滤波原理，提出一种基于地物光谱矢量特征的滤波去噪方法MFS，用定标后的Landsat-7ETM+地物反射率图像广义归一光谱进行滤波处理实验，保持图像地物光谱特征、边沿特征、纹理特征、地形因子、地物BRDF因子、混合像元每种地物占比因子前提下，消减了噪声，提高了像元值精度及图像清晰度。 MFS不需要DTM数据能适应地形变化，山区图像去噪与平原区等效、暗区图像去噪与亮区等效、全景图像去噪效果均衡一致。MFS维持原图像物理量纲不变，去噪同时增强地物信号，，提高了图像信噪比，适用于遥感图像预处理。全文共分为五章，第一章为绪论，主要介绍研究背景、国内外研究现状、遥感图像去噪的经典算法以及研究背景与意义。第二章首先介绍地物反射特征的描述方法，然后详细论述归一光谱矢量理论以及由归一光谱矢量推出的广义归一光谱矢量理论，最后介绍了MFS滤波模型的算法。第三章主要介绍本文滤波处理实现的开发语言C#以及根据该理论所进行编程实现的处理程序。第四章主要是结合前面几章的内容，以Landsat-7ETM+遥感图像为实验数据，对ETM+第1波段进行去噪处理，与目前常用的4种滤波方法进行比较，MFS去噪效果优势比较明显，可望取代遥感图像现行去噪方法，有一定应用价值。第五章为本文的结论，主要是对本文所取得的进步与存在的问题以及对下一步工作的展望。
[Abstract]:The remote sensing images obtained by ground observation of ETM remote sensor have high frequency components of atmospheric interference, residual errors left by image processing in the early stage and other errors with unknown sources, so it is necessary to Denoise in practical applications. In order to improve the clarity of the image, scholars have explored many methods, among which the low-pass filtering effect is better, which is widely used by remote sensing image processing software, but the filtering effect is still not ideal, the interior of the ground object is weak and clear, and the edge of the ground object is weakly visible and gelatinized. The noise is weakened and the signal is weakened at the same time. The denoising method needs to be improved and the new method needs to be explored. It is hoped that the new method can weaken the noise, enhance the signal and output the ground object image with high accuracy and clarity. In this paper, based on the principle of low-pass filtering, a filtering denoising method based on the spectral vector characteristics of ground objects is proposed. MFS, carries on the filtering experiment with the generalized normalization spectrum of the calibrated Landsat-7ETM figure reflectivity image to maintain the spectral characteristics of the image ground objects. On the premise of edge feature, texture feature, terrain factor, ground object BRDF factor and mixed pixel ratio factor, the noise is reduced and the pixel value accuracy and image clarity are improved. MFS does not need DTM data to adapt to terrain changes, mountain image denoising is equivalent to plain area, dark area image denoising is equivalent to bright area, panoramic image denoising effect is the same. MFS keeps the physical dimension of the original image unchanged, denoising and enhancing ground object signal. The signal-to-noise ratio (SNR) of the image is improved and it is suitable for remote sensing image preprocessing. The full text is divided into five chapters. The first chapter is the introduction, which mainly introduces the research background, the research status at home and abroad, the classical algorithm of remote sensing image denoising, as well as the research background and significance. In the second chapter, the description method of reflection characteristics of ground objects is introduced, and then the theory of normalized spectral vector and the theory of generalized normalized spectral vector derived from normalized spectral vector are discussed in detail. Finally, the algorithm of MFS filtering model is introduced. The third chapter mainly introduces the development language C # and the programming program according to the theory. The fourth chapter mainly combines the contents of the previous chapters, takes Landsat-7ETM remote sensing image as the experimental data, carries on the denoising processing to the first band of ETM, compared with the four filtering methods commonly used at present, the MFS denoising effect is more obvious. It is expected to replace the current denoising method of remote sensing image and has certain application value. The fifth chapter is the conclusion of this paper, mainly on the progress and existing problems of this paper, as well as the prospect of the next work.
【学位授予单位】：东北师范大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP751

【参考文献】