卫星多源遥感图象象素级融合技术研究

【作者】 刘哲；

【导师】 郝重阳；

【作者基本信息】 西北工业大学 ， 信号与信息处理， 2003， 硕士

【摘要】 多源遥感图象数据融合(也称信息融合)是近年来遥感领域和图象工程领域的研究热点。本论文针对多光谱图象传感器数据与全色图象传感器数据进行双源融合。利用多光谱图象高光谱信息与全色图象高空间信息两者之间的互补性，研究在尽可能保留原光谱信息的前提下，提高空间分辨率的象素级图象融合算法。 论文在研究数据融合的基本概念和基本原理及遥感图象融合预处理的过程与步骤的基础上，具体分析了象素级融合的常用方案，及其相关的IHS，YIQ，PCA，HPF四种传统的经典算法。同时，在研究主客观两方面融合结果评价标准的前提下，比较了多波段融合中各种算法的优缺点，以及单波段融合法中HPF(高通滤波)法与小波变换法的异同。作者在重点研究基于小波变换融合算法的过程中，取得了如下一些具有新意性的成果： 1．提出了一种新的基于特征量积的融合判决依据。这种特征量积是在经典的IHS彩色空间变换和小波多分辨率分析的基础上，利用图象高频小波系数的多个特征量来定义的。 2．在选定特征量积作为判决依据的基础上，用“选择性叠加融合”替代了常见的“选择性融合”的融合思路。 3．将上述两者综合运用，形成了一种图象象素级融合新算法。 通过一组多光谱图象和全色图象数据进行了融合仿真试验，将本文算法与IHS、HPF等传统经典算法和归一化矩算法的试验结果作了比较。结果表明了本文算法的有效性，并且能在保留多光谱图象光谱信息的同时，使多光谱图象的空间分辨率有了提高。更多还原

【Abstract】 In recent years, multisensor image fusion has attracted many attentions in remote sensing area and image project area. This paper focuses on image fusion between multispectral images and panchromatic images. In order to enhance the result image’s spatial resolution with preserving spectral information, research on algorithms is developed in next chapters.This dissertation firstly studies the cardinal concepts and theories of multisensor data fusion (information fusion). At the same time, this paper introduces steps and procedures of pretreatment in remote sensing image fusion. Afterwards, the paper systematically research the techniques used in pixel-level image fusion and four traditional algorithms (IHS, YIQ, PCA, HPF). Based on making out evaluation criteria of result image in impersonality and personality, merits and demerits of algorithms on multi-band image fusion are discussed. At the same time, HPF(high pass filter) method is compared with wavelet method among single-band image fusion methods. Author stresses the research on the details of algorithm based on wavelet transform, and gets such fresh ideas as:1. This paper puts forward a new measurement based on feature product. Feature product was defined with feature moments of the wavelet coefficients with high frequency component under different resolutions, based on transform of IHS and wavelet multiresolution analysis.2. Based on the measurement of feature product, a new idea that additive select fusion take the place of select fusion is introduced.3. According to the local feature product and the new idea ,a new pixel-level image fusion algorithm is produced.At last, computer simulations of image fusion are carried out by a group of multispectal images and panchromatic images. Compared with the result of IHS , HPF traditional fusion methods and the normalized correlation moment method, it is proved the validity of this method ,which can improve spatial resolution of the result images with preserving spectral information.更多还原