基于各向异性双变量收缩的含噪图像融合

针对含噪图像的融合问题，基于各向异性双变量收缩提出了一种图像融合的新算法。在小波域内，对高频子带进行各向异性双变量收缩以压缩噪声，对收缩系数进行模值选大融合，对低频子带进行算术平均融合，对融合系数进行逆变换获得融合图像。对多聚焦图像及红外与可见光图像的融合结果进行主客观性能评价表明，所提出的方法可以更好地保留源图像的细节信息，并有效抑制噪声。

Noisy image fusion based on anisotropic bivariate shrinkage

A new image fusion method based on anisotropic bivariate shrinkage is proposed, focusing on the fusion of noisy images. The high frequency subbands are shrunk with anisotropic bivariate shrinkage in the wavelet domain to sup-press the noise effectively. The shrunk coefficients are fused by maximum absolution selection while average is taken to fuse the low frequency subbands. The fused image is obtained by taking inverse transform of the fused coefficients. Both subjective and objective assessments of the fusion results of multi-focus images and IR and visible images show that the proposed method works better in preserving details and suppressing noises of the source images.