基于离散余弦变换和聚类算法的遥感图像融合

Remote Sensing Image Fusion Based on DCT and Clustering Algorithm

本文将二维离散余弦变换(DCT)应用于多光谱图像与全色光学图像融合问题的研究中。首先,对全色光学图像和多光谱图像的亮度分量(Y)作DCT变换;根据变换矩阵DCT系数分布特点,对DCT系数进行聚类,通过统计不同类的DCT系数的数目,将DCT变换矩阵分为低频分量和高频分量;用Y分量的低频分量代替全色光学图像的低频分量,保留全色光学图像的高频分量,反DCT变换,得到最终的融合结果。实验表明,融合图像在空间细节信息的表现能力和光谱特性保持性能上均是比较优良的,说明该方法是可行有效的。

This paper applies 2-D Dcr to the remote sensing image fusion between panchromatic and multi-spectrul images. First, perform a DCT transform on the panchromatic image and Y component respectively, which is obtained by performing a YIQ transform on the multi-spectral image. According to the DCT coefficient＇s distribution property of the panchromatic image, classify DCT coefficient into several classes, and divide DCT coefficient into the component of low frequency and the component of high frequency by calculating the number of the different class. And then substitute the low frequency component of the multi-spectrul image for the one of the panchromatic image, and preserve the high frequency component of the panchromatic image. Finally, perform an inverse DCT transform and the fused image will be gained. Experiments show that the fused image can reach a trade-off between spectral and spatial information well and the approach is feasible and efficient.