摘要
由于omis影像128个波段间具有较强的相关性,可对高光谱响应曲线采用小波分解,用噪声较小波段的部分高频信息代替噪声较大波段的相应高频信息对其进行小波降噪。再对降噪后的影像采用db4函数进行7级小波分解,对各级小波系数取一范数后,用最小二乘法对各级分解对应的范数在半对数坐标系下线性拟合,根据直线斜率求取各像元分维值,最后实现分类。通过实验,证明了此方法的有效性。
As the omis image＇s 128 bands had a strong inter-band correlation, wavelet decomposition can be used for high spectral response curves. Took high-frequency messages of the low-noisy bands to replace the corresponding high-frequency information of high-noisy bands to reduce noise by wavelet. In this paper, after noise reducing, 7 degrees decompositions of pixel spectral curve, based on db4 function, will be used, then solved the first norm of wavelet coefficients at all levels and fit them to its decomposing scales under the method of least squares linear in the semi-logarithmic coordinates, later get fractal dimensions according to the slope of this line, according to which a final classification image will be achieved. At last, the experiments confirmed the effectiveness of this method.
出处
《地理空间信息》
2010年第5期89-91,共3页
Geospatial Information
基金
国家重点基础研究发展计划资助项目(2006CB701303)
关键词
高光谱遥感影像
小波降噪
小波变换
分形维
hyperspectral remote sensing images
noise reduce by wavelet
wavelet transform
fractal dimension
