基于改进的非下采样剪切波变换多聚焦图像融合技术的研究

Research of Multi-focus Image Fusion Technology Based on Improved Non-subsampled Shearlet Transform

针对多聚焦图像融合的具体问题,文中提出一种基于改进的非下采样剪切波变换(NSST)的图像融合方法。NSST变换更加有利于保持图像的边缘和轮廓信息,同时保持了图像的平移不变性。在对原图像进行多尺度几何变换后,针对图像融合过程中源图像不同清晰指标,采用改进的绝对值取大的融合规则处理经过多尺度几何变换后的高频系数;采用基于区域加权的拉普拉斯能量和的方式处理低频系数,将得到的高、低频系数经过NSST逆变换最终得到融合图像,实验结果表明,对于多聚焦图像融合,文中提出的算法,不仅在主观视觉方面获得了良好的效果,而且在客观评价标准方面也优于传统的多聚焦融合算法。

In order to solve the problem of multi focus image fusion,a new image fusion method based on non-sampling shearlet transform is proposed in this paper. The NSST transform is more conducive not only to keep the image edge and contour information,but also to maintaining the translation invariance. In multi-scale geometric transform of the original image,for the image fusion process in different source image clear indicators,the absolute value of the improved fusion rule was used in the high frequency coefficients of multi-scale geometric transform. While for the low frequency coefficients,the area weighted method was used based on the Laplasse energy. After the high and low frequency coefficients after NSST transform,the final fusion image was obtained. The experimental results show that the multi focus image fusion obtained by the algorithm proposed in this paper not only has good effect in the subjective aspect,but also superior in the focus objective the evaluation standard of the traditional fusion algorithms.