一种基于视觉显著性的图像融合算法

Image Fusion Method Based on Visual Saliency

基于视觉显著特征提出一种新的图像融合算法。首先对源图像进行小波分解,得到低频系数和高频系数;然后针对低频系数提出一种基于视觉显著性的融合规则,对高频系数采用绝对值取大的融合规则;最后对融合后的低频系数和高频系数进行小波逆变换得到最终融合图像。该方法能完好地将源图像的细节融合在一起。仿真实验表明,该算法在视觉效果上比传统及改进方法更好,同时互信息、平均结构相似性、信息熵等传统客观评价指标有所提高。

This paper proposed a new approach for image fusion based on visual saliency .Sourse images are decomposed by wavelet to get coefficients of low and high frequency,then we fuse part of low frequency using salient method proposed and fuse part of high frequency by selecting arge bsolute avelet coefficients . Fused image will be got by wavelet inverse transformation.This method will fuse details of source images beautifully .It’s proved that method proposed outperforms the traditional and improved method in visual effect and some assessment criteria are advanced ,such as mutual information, average structural similarity ,entropy and so on.