In this paper, we propose an image filter based on fuzzy gradient threshold function and global optimization, termed as gradientlet filter, from the perspective of luminance and gradient separation. It can remove small gradient textures and noise while maintaining the overall brightness and edge gradients of an image. Based on gradientlet filter and image saliency, we further put forward a new method for infrared and visible image fusion, which can overcome the challenges of low contrast, edge blurring and noise existing in traditional fused images. First, the gradientlet filter is used to decompose source images into approximate layers and residual layers, where the former reflects the overall brightness of source images without edge blurring and noise, and the latter reflects the small gradient texture and noise of source images. Second, according to the characteristics of the approximate and residual layers, we propose contrast and gradient saliency maps and construct corresponding weight matrices. Finally, the fused image is obtained by fusion and reconstruction based on previously obtained sub-images and weight matrices. Extensive experiments on publicly available databases demonstrate the advantages of our method over state-of-the-art methods in terms of maintaining image contrast, improving target saliency, preventing edge blurring, and reducing noise.

本文从亮度和梯度分离的角度提出了一种基于模糊梯度阈值函数和全局优化的图像滤波器，称为梯度波滤波器。它可以去除小的渐变纹理和噪点，同时保持图像的整体亮度和边缘渐变。基于梯度波滤波和图像显着性，我们进一步提出了一种红外与可见光图像融合的新方法，可以克服传统融合图像存在的对比度低，边缘模糊和噪声大的问题。首先，使用梯度波滤波器将源图像分解为近似层和残差层，其中前者反映源图像的整体亮度，而没有边缘模糊和噪声，而后者则反映源图像的小梯度纹理和噪声。第二，根据近似层和残差层的特征，提出对比度和梯度显着图，并构造相应的权重矩阵。最后，基于先前获得的子图像和权重矩阵，通过融合和重建获得融合图像。在可公开获取的数据库上进行的大量实验证明，在保持图像对比度，提高目标显着性，防止边缘模糊和减少噪声方面，我们的方法相对于最新方法的优势。