结合区域特性的有限离散剪切波图像融合

Image fusion based on area characteristics in finite discrete shearlet domain

为了提高多聚焦图像和红外与可见光图像的融合精度,结合有限离散剪切波变换具有良好的局部化特性及平移不变性,提出了一种基于有限离散剪切波变换(FDST)的图像梯度信息相关性因子加权与对比度相结合的融合算法。首先对严格配准后的图像进行FDST分解,得到低频子带系数和不同尺度不同方向的高频子带系数;然后对低频子带系数采用图像梯度信息相关性因子加权融合算法,高频则利用对比度将低频系数与高频系数联系起来并以对比度作为度量系数取舍的准则进行融合;最后应用有限离散剪切波逆变换重构得到融合图像,并对融合结果进行主观视觉和客观评价。实验结果表明,该算法在主观视觉效果和客观评价指标上优于其它融合算法。

In order to improve the accuracy of multi-focus image and infrared visible image fusion, based on the advantages of the good localization and shift invariance of the finite discrete shearlet trans- form, we propose a new image fusion algorithm, called FDST, which combines contrast with image gra- dient information correlation factor. Firstly, the FDST decomposes the registration images, and obtain the low frequency sub-band coefficients and high frequency sub-band coefficients of different scales and directions. The fusion principle of low frequency sub-band coefficients bases on the method of image gra- dient information correlation factor for weighting. The high frequency sub-band coefficients combine the high frequency and low frequency sub-band coefficients with contrast, and uses contrast as the criterion for choosing metric coefficients which is taken as the fusion rule. Finally, the low frequency information and high frequency information are reconstructed to obtain a fused image by the finite discrete shearlet inverse transform, and both subjective visual evaluation and objective performance assessments of the fusion results are implemented. The results indicate that the algorithm is superior to other fusion algo- rithms in subjective visual effects and objective evaluation.