基于Shearlet变换的自适应图像融合算法

针对多聚焦图像与多光谱和全色图像的成像特点,结合Shearlet变换具有较好的稀疏表示图像特征的性质,提出了一种新的图像融合规则.并基于此融合规则,提出了基于Shearlet变换的自适应图像融合算法.在多聚焦图像的融合算法中,分别对聚焦不同的图像进行Shearlet变换,并基于本文提出的融合规则,对分解后的高低频系数进行融合处理. 通过与多种算法的比较实验证明了本文提出的算法融合的图像具有更高的清晰度和更加丰富的细节信息.在多光谱和全色图像的融合处理中,提出了一种基于Shearlet变换与HSV变换相结合的图像融合方法.该算法首先对多光谱图像作HSV变换,将得到的V分量与全色图像进行Shearlet分解与融合,在融合过程中对分解系数选用特定的融合准则进行融合,最后将融合生成新的分量与H、S分量进行HSV逆变换产生新的RGB融合图像. 该算法在空间分辨率和光谱特性两方面达到了良好的平衡,融合后的图像在减少光谱失真的同时,有效增强了空间分辨率. 仿真实验证明,本文算法融合的图像与传统的多光谱和全色图像融合算法相比,具有更佳的融合性能和视觉效果.

Adaptive Image Fusion Algorithm Based on Shearlet Transform

For the imaging characteristics of multi-focus images, multi-spectral images and panchromatic images, since the Shearlet transform has better properties to sparse express the characteristics of the images, a kind of new image fusion rules is proposed. Moreover, based on the fusion rules, the algorithm of adaptive fusion rules based on Shearlet transform is proposed. In the algorithm of multi-focus images fusion, the different focus images are transformed with Shearlet transform respectively, and the decomposed low-frequency coefficients and high-frequency coefficients were fused according to the proposed fusion rules. It is verified that the proposed algorithm has better clarity and richer details information compared with many algorithms. Multi-spectral and panchromatic images fusion algorithm is proposed based on combination of Shearlet and HSV transform. Firstly, the multi image is transformed with HSV transform; then, the gotten V component is Shearlet transformed with pan image and the specific fusion rules is chosen for the decomposition coefficient in the fusion process; finally, the new V and H,S components are transformed with inverse HSV transform. This algorithm reached a good balance in the two aspects of spatial resolution and spectral characteristics. The fused images can reduce spectral distortion, and effectively enhance the spatial resolution. The simulation experiments show that the proposed algorithm has better fusion performance and visual effect, compared to traditional multi-spectral and panchromatic images fusion algorithms.