Contourlet变换低通滤波器对图像融合算法影响的讨论

根据Contourlet变换多尺度、多方向的特性以及自然图像的频谱特点, 分析了Contourlet变换低通滤波器对图像融合算法性能和融合结果中伪轮廓的影响, 讨论了低通滤波器与分解层数选取之间的关系. 理论与实验证明, 使用带宽较窄的低通滤波器能够得到较好的融合结果, 当低通滤波器系数均为正数时融合图像的伪轮廓显著减少. 本文比较了使用不同低通滤波器和不同融合准则时得到的融合结果, 实验结果证明, 使用合适的低通滤波器, 可以在选用较简单的融合准则和分解较少层数的情况下, 得到很好的融合结果, 大大减小了融合算法的复杂度.

Discussion upon Effects of Contourlet Lowpass Filter on Contourlet-based Image Fusion Algorithms

According to multiscale and multidirection contourlet transform and considering real image frequency spectrum characteristics, effects of contourlet lowpass filter on contourlet-based image fusion algorithms and ringing artifacts in fused images were analysed, and relations between lowpass filter and decomposing levels were discussed. Theories and experiments proved that better fusion results can be obtained with a narrow bandwidth lowpass filter, and that ringing artifacts can be eliminated by using a lowpass filter with positive coefficients. Fusion results using different lowpass filters and fusion rules were compared. Experimental results showed that choosing an appropriate lowpass filter can get a satisfying fusion result, with an easier fusion rule and less decomposing levels, which can reduce the fusion algorithm's complexity.