针对单一能量不能完整体现复杂结构件数字射线成像的全部图像信息这一问题,提出了一种基于引导滤波、加权最小二乘滤波(GF-WLS)和视觉显著图(VSM)的图像融合方法。首先,采用两种滤波方式对图像进行多尺度细节提取,提取出的细节作为细节层,原图作为基础层;其次,根据相位一致性和像素对比度原理分别对细节层和基础层构造视觉显著图,进而通过像素显著性的对比得到初始权重图,再利用引导滤波去除噪声;然后,将权重图归一化作为加权映射权重分别指导基础层与细节层的融合;最后,将融合后的两者进行叠加得到融合结果。试验表明,所提方法能够获得更高质量的融合图像。

To address the problem that a single energy cannot completely reflect all the image information of digital radiography of complex structural parts, we propose an image fusion method based on guided filtering, weighted least squares filtering (GF-WLS) and visual saliency map (VSM). Firstly, two filtering methods were used for multi-scale detail extraction of the image, with the extracted details being used as the detail layer and the original image as the base layer. Secondly, the visual saliency map was constructed by using phase consistency(PC) and pixel contrast. The initial weight map was obtained by the comparison of pixel saliency, and then the noise was removed by using the guided filter. Thirdly, the weight map was normalized as the weighted mapping weights to guide the fusion of the base layer and the detail layer. Finally, the fusion results were obtained by superimposing the two. The experimental comparison showed that the proposed method was able to obtain higher quality fused images.