Abstract: This paper proposes a fusion algorithm for the color image?and near infrared image. Firstly, a bilateral filter is proposed which has an improve kernel function. Secondly this scheme implements bilateral filter and Gauss filter for color image and near infrared image, to get multi-scales images, for example the low-frequency image, detail image, edge image. Thirdly, an adaptive weight selection method is proposed when different scale images are incorporated. The method considers the neighborhood of each pixel which is used to fusion, when images are combined. It makes the fusion images to reduce noises and color of the fusion image to be closer to the original image. The experiments show the obtaining fusion image has color information which accords with human observations habits. Otherwise, its information of the texture, edge details is more abundant than the original color.

Key words: bilateral filter, image fusion, kernel function, adaptive, near field

摘要： 提出一种彩色图像和近红外图像的融合算法，先给出一种改进核函数的双边滤波器，然后应用该双边滤波器和高斯滤波器相结合多次作用于彩色图像和近红外图像，得到低频图像、边缘图像、细节图像等不同尺度的图像，在不同尺度图像融合的权重选择上，提出一种自适应的权重选择法。该方法综合考虑参与融合的各像素点的邻域来确定各点权重，不仅抑制了融合图像噪声，还使融合图像色彩更贴近原彩色图像。实验表明该方法得到的融合图像具有符合人眼观测习惯的颜色信息，此外融合图像的纹理、边缘等细节信息比原彩色图像更丰富。

关键词: 双边滤波, 图像融合, 核函数, 自适应, 邻域