基于多分辨率分割的区域图像融合

An Image Fusion Method Based on Multi-Resolution Segmentation

图像融合是一项综合同一场景的多幅源图像信息的技术。现有的区域图像融合方法或者是只对最高层低频带分割并以此分割信息来指导所有层的融合,或者是其多分辨率分割方法过于复杂难以满足实时性。鉴于此,该文发展了一种基于多分辨率分割的区域图像融合方法。它的主要特点是多分辨率分割。其步骤为:首先对源图像进行小波变换的多分辨率分解,然后对分解后每一层的低频图像都进行区域分割,最后用每一层分割得到的区域信息来分别指导每一层的融合。仿真表明该文发展的基于多分辨率分割的区域图像融合方法的融合性能要优于传统的基于窗口的图像融合方法和只对最高层低频带分割的区域图像融合方法。

Image fusion method is a technology for synthesizing the information coming from multi original images of the same scene. Present region - based fusion methods can be classified into two classes ： one class is to segment low - low frequency part of the highest layer to instruct the fusion of all layers; the other is multi - resolution segmentation method which is too difficult to meet the real -time requirement. Therefore, this paper develops a multi -reso- lution segmentation based fusion method. Its main characteristic is multi resolution segmentation. Its steps are： firstly it decomposes the original images by multi - resolution based on wavelet transform; secondly it segments low - low frequency part of each layer into different regions; finally it uses the region information of each layer to instruct the fusion of each layer, respectively. Simulation experiment indicates that the performance of this multi - resolution segmentation based image fusion method is better than those of the ordinary window - based fusion method and of the region - based fusion method only for segmentation of low - low frequency part of the highest layer.