关键词: 多波段图像配准, 快速归一化空间频谱互相关, 归一化互相关, 模板匹配

Abstract: Among the methods for image registration, the normalized cross correlation (NCC) method is the most widely used one, which is conceptually straightforward and easy to implement. The classic NCC method is based on spatial domain, works for single band image and does not leverage the spectral information of all spectral bands of images. A novel normalized spatial-spectral cross correlation (NSSCC) method was proposed recently by some researchers, the NSSCC method utilizes all spectral bands for multi-spectral image registration. This NSSCC method effectively increases the registration reliability and discrimination compared to the classic NCC method. However, if the image contains many spectral bands and its size is large, the NSSCC method will require high computation cost. This paper presents an improved algorithm for fast calculation of the NSSCC method and its application to the problem of multi-spectral image registration. The simulation results show that the improved algorithm can effectively reduce the computation cost of the NSSCC method.

Key words: multi-spectral image registration, fast normalized spatial-spectral cross correlation, normalized cross correlation, template matching