LNSST域灰度突变度的红外与可见光图像融合

LNSST Domain and Mutation Degree of Greyscale of Infrared and Visible Light Image Fusion

对红外与可见光图像融合过程中出现的融合时间较长、细节信息提取不足的缺点,提出一种局部化非降采样剪切波变换(Local Non-subsampled Shearlet Transform,LNSST)的红外与可见光图像融合提升算法。对红外图像和可见光图像采用LNSST域算法进行多层次、多方向的分解,得到低频子带系数和各带通方向子带系数。低频部分使用基于灰度突变度的权值平均的融合规则;高频部分采用区域纹理平滑度和拉普拉斯能量和相结合的融合方法,然后对低频融合系数和高频融合系数进行LNSST逆变换得到融合后的图像。用不同的算法进行验证结果显示本文提出的融合算法生成的融合图像细节更多、更清晰,执行时间相对较短。

Aiming at the shortcomings of long fusion time and the details not fully extracted in the infrared and visible image fusion, a novel image fusion algorithm based on local non-subsampled Shearlet transform is proposed. LNSST is used to decompose the source image from multi-direction and multi-scale. The low-frequency components and every bandpass subband direction components are obtained. For the low frequency subband components, the fusion method of the weighted average of gray mutation degree is adopted; for the top of the high-frequency components, the fusion method of regional texture smoothness and the Laplace energy is adopted. Finally, we combine the fusion low frequency subband fusion coefficients and the bandpass subband direction fusion coefficient to execute the LNSST inverse transformation to get the final fusion images. The results show that the fusion algorithm generated by the fusion algorithm is clearer and the execution time is relatively short.