基于混合模型驱动的红外与可见光图像融合

为了解决红外与可见光图像融合中显著特征不突出、图像对比度低的问题,提出一种混合模型驱动的融合算法.首先,采用潜在低秩表示模型分别提取红外与可见光图像的基础子带、显著子带及稀疏噪声子带;其次,采用非下采样剪切波变换模型将基础子带分解为低频系数和高频系数,对低频系数采用字典学习和稀疏表示进行精确拟合,对高频系数采用局部窗口结合逻辑加权进行选择;再次,显著子带采用区域能量比阈值自适应加权法进行融合;最后,对融合后的低频系数和高频系数进行一级重建,得到融合基础子带,舍弃稀疏噪声子带,再结合融合显著子带进行二级重建,得到融合图像.实验结果表明:所提出算法能够得到蕴含丰富信息且较为清晰的融合图像,具有可行性;融合结果的对比度较高,目标轮廓显著,能够提升场景的辨识度,具有有效性.

Infrared and visible image fusion based on hybrid model driving

In order to solve the problems of less-prominent features and low image contrast in infrared and visible image fusion, the paper proposes a hybrid model-driven fusion algorithm. Firstly, the latent low-rank representation model is used to extract the base sub-band, significant sub-band and sparse noise sub-band of the infrared and visible image respectively. Then, the base sub-band is decomposed into low frequency coefficients and high frequency coefficients by the non-subsampled Shearlet transform model. The low frequency coefficients are accurately fitted by dictionary learning and sparse representation, while the high frequency coefficients are selected by local window combined with logic weighting. The significant sub-band is fused by adaptive weighting of regional energy ratio threshold. Finally, the fused low frequency coefficients and high frequency coefficients are $1^st$-order reconstructed to obtain the fused base sub-band and discard the sparse noise sub-band. Then the $2^nd$-order reconstruction is conducted with the fused significant sub-band to get the fused image. The experiment results show that the proposed algorithm can obtain clear fused image with rich information. The fused results have high contrast and significant target contour which can improve the recognition of the scene.