基于均匀离散曲波变换的多聚焦图像融合

利用均匀离散曲波变换（UDCT）多尺度、多方向、低冗余等特征，提出了一种新的多聚焦图像融合方法。首先使用UDCT对源图像进行多频带分解；然后根据多聚焦图像的特点，对分解后的低频子带系数运用一种基于改进拉普拉斯和算子的方案进行融合，对高频方向子带系数运用基于局部能量的融合规则进行融合，并对融合系数做一致性检测；最后重建各子带系数得到融合图像。实验结果表明:所提方法可以有效地融合源图像中的方向信息和细节特征，同时抑制了融合图像中的伪Gibbs现象；与基于拉普拉斯金字塔分解、小波变换以及轮廓波变换的图像融合方法相比，该方法取得了更好的视觉效果和量化结果。

Uniform discrete curvelet transform for multi-focus image fusion

A novel fusion algorithm for multi-focus images was proposed using Uniform Discrete Curvelet Transform(UDCT) for its characteristics of multi-scale; multi-direction and low redundancy. First, the source images were decomposed into several subbands using UDCT. Then, according to the characteristics of multi-focus images, the coefficients of low-frequency subband were fused with a scheme based on the sum-modified-laplacian; the coefficients of high-frequency subbands were fused with the fusion rule based on local energy; and the consistency of the fused coefficients was verified. Finally, the subband coefficients were reconstructed, and the fused image was obtained. The experiment results indicate that the proposed method can effectively fuse the directional information and detailed features of source images, and suppress pseudo-Gibbs phenomena of fused image; compared with other fusion methods, such as those based on Laplacian pyramid transform, discrete wavelet transform and contourlet transform, this method obtains better fusion quality in terms of both visual and quantified measure.