Application of Triangulation-Based Image Registration Method in the Remote Sensing Image Fusion

Abstract

Based on image registration principle of the feature point, this article brings forward methods of a combination of Moravec and Forstner operator, extracts feature points in the base image making use of sequential detection algorithm from coarse to fine, and realizes automatic matching of the same name point in the pending registration images through the geometric distance constraint conditions. After the triangulation through the triangulation methods in this article, the images and the results by the second transformation method have similar performance, but in the process of registration it can save a lot of time and energy. Count the three aspects of information including entropy, degrees of spectral distortion and into-degrees of high-frequency information in the fused image. Of which entropy expresses the amount of images information. The value of entropy in the fused image is larger than the original panchromatic image and multi-spectral image, accounting for through the image fusion it indeed improves the information content of the image! In Shenzhen City Digital City project, the results by using of the method showed that the fusion effect is significantly higher than other traditional methods.