结合小波变换的梯度场图像渐进融合算法

Gradient Field Image Progressive Fusion Algorithm Combined with Wavelet Transformation

基于梯度场的图像融合算法只适用于尺度差异不大于1∶4的多光谱图像与全色图像。针对尺度差异为1∶8的北京一号卫星多光谱图像及高分辨率全色图像的融合问题,提出一种结合小波变换的梯度场图像尺度渐进融合算法。利用小波变换方法将多光谱图像与高分辨率全色图像尺度差异倍数缩小,得到基于小波变换的初级融合,再进行基于梯度场的Poisson图像融合。实验结果表明,渐进融合图像与多光谱图像的平均颜色差异值为23.5,与高分辨率全色图像的平均梯度差异值为2.1,多尺度纹理特征值差异值分别为3.98、10.2、18.9,渐进融合图像与高分辨率全色图像的空间细节和纹理细节吻合程度更好。

Image fusion algorithm based on gradient field is one of relatively new remote sensing image fusion algorithms. But the fusion algorithm is only suitable for using in less than 1：4 scale ratio of multi-spectral image and panchromatic image. In order to solve the fusion problem, which is caused by the scale differences between Beijing-1 satellite multi-spectral image and panchromatic image as 1：8, this paper presents an image progressive fusion algorithm based on wavelet transformation and gradient field. It uses wavelet transformation to narrow the scale difference of multi-spectral images and high-resolution panchromatic images, and gets the preliminary fusion image by wavelet transform fusion algorithm, and makes preliminary fusion image and high-resolution image fusion based on gradient field. Experimental results show that the average color difference between the progressive fusion image and the multi-spectral image is 23.5. The average gradient difference and the multi-scale texture value differences between the progressive fusion image and the high-resolution panchromatic image are 2.1, 3.98, 10.2, 18.9, that is better than the other mainstream fusion algorithm, which indicates that the matched degree of the texture details between the progressive fusion image and the high-resolution panchromatic image is much better.