基于归一化相关矩的多分辨率遥感图象融合

多传感器数据融合技术已广泛应用于遥感图象处理方面. 针对遥感多光谱图象空间分辨率较低的问题,提出了一种基于归一化相关矩的多分辨率图象融合方法. 该方法首先对图象进行二维小波变换, 然后根据所得到的高频小波系数的一阶、二阶统计特征来定义图象局部灰度相关矩, 并以此作为图象融合测度来对遥感图象进行多分辨率特征融合, 从而得到包含更多信息和有效特征的融合图象. 仿真结果表明, 融合后的图象在保留多光谱信息和提高空间分辨率上均能获得较好的效果, 因而可以更好地用于目标识别、分类等遥感图象处理方面.
关键词　数据融合　多分辨率图象融合　相关矩　遥感图象处理
Abstract　M ult isenso r data fusion technique has been w idely used in remo te sensing image p rocessing. A majo r p roblem in using remo te sensing mult ispect ral images is the low spat ial reso lut ion. In o rder to imp rove spat ial reso lut ion of the mult ispect ral images, a mult ireso lut ion image fusion method based on the no rmalized co rrelat ion moment is p ropo sed in th is paper. A fter 22D w avelet t ransfo rm the no rmalized co rrelat ion moment is defined by means of stat ist ic first2o rder and second2o rdermoment of tho sew avelet coefficientsw ith h igh frequency component under different reso lut ions. By using the local co rrelat ion moment asmeasure of feature select ion, the new method carries out mult ireso lut ion images fusion. A nd the fused images include mo re useful info rmat ion fo r further
app licat ion. Experiments are conducted on L andsat TM and SPO T data. Computer simulat ions show that the fused images obtain bet ter results in term s of bo th p reserving spect ral info rmat ion and imp roving spat ial reso lut ion of the mult ispect ral images, and are rather used to many aspects of remo te sensing image p rocessing such as target recognit ion, ground materials classificat ion, etc.
Keywords　Data fusion, M ult ireso lut ion image fusion, Co rrelat ion moment, Remo te sensing image p rocessing