A New Image Fusion Algorithm Based on Classification
一种基于分类的融合算法

This paper presents a new classification-based fusion scheme, which is used throughout this paper as a synonym of the "pixel-level fusion". It can be applied to merge low-resolution images and co-registered high-resolution images. The key of this trial is as follows: (1) Geometrical co-registration IKONOS 1-m panchromatic image and 4-m multi-spectral image; (2) Classification of the High-resolution Panchromatic image together with the low-resolution images using supervised or unsupervised algorithms; (3) According to each class histogram of High-resolution image, adjustment of the corresponding spectral values of single multi-spectral image; (4) HIS transformation of the adjusted multi-spectral image using cylindrical coordinates, and acquirement of the fusion images. From this algorithm, we can conclude as follows: (1) The proposed method can merge the class information, spatial information of high resolution image and spectral information of Low-resolution image, accounting for the importance of classification as prior knowledge ; (2) Based on precise classification, it can effectively eliminate the false color at the edge of objective and have better visual effects; (3) It can try to add classification and spatial relationship to image fusion. In this trial the selected Peking image of IKONOS will be used, which can bring us convenience when verifying our algorithm.