基于简化脉冲耦合神经网络与拉普拉斯金字塔分解的彩色图像融合

Color image fusion based on simplified PCNN and Laplace pyramid decomposition

针对图像空间分辨率低及分类精度不高等问题,提出一种基于简化脉冲耦合神经网络(S-PCNN)与拉普拉斯金字塔分解算法的彩色图像融合算法。首先,把RGB图像转换到HSI彩色空间中得到H、S、I三个分量,将H分量输入到S-PCNN模型中,利用S-PCNN对H分量进行特征区域聚类后,基于脉冲震荡频图和局部熵实现各源图像的H分量融合;然后采用拉普拉斯金字塔对S、I分量进行分辨率分解,根据不同融合策略对不同拉普拉斯金字塔图层中的S、I分量进行融合。最后,对融合后的H、S、I分量进行彩色空间逆变换,得到最终的RGB图像。实验结果表明,该融合算法在清晰度、空间频率、标准差方面优于传统的主成分分析(PCA)、脉冲耦合神经网络(PCNN)等算法,能很好地保留源图像的细节、纹理和主要特征信息,有效地提高了图像的融合效果。

A color image fusion algorithm based on Simplified Pulse Coupled Neural Network（ S-PCNN） and Laplace pyramid decomposition algorithm was proposed to solve the problem of low spatial resolution and low classification accuracy.First of all,the RGB image was translated into HSI color space to get three components of H,S,and I,then input the H component into S-PCNN.By using S-PCNN to carry out on the H component feature area clustering,oscillation frequency graph and local entropy of the pulse were used to realize the H component fusion of each source image;and then the resolution of the S and I components were decomposed using Laplace pyramid.After that,the S and I components in different Laplace pyramid layers were fused according to the different fusion strategies.Finally,the RGB image was obtained through inverse transform of the fused components of H,S and I in the color space.Experimental results show that the proposed algorithm is better than the traditional ones such as PCA（ Principal Componet Analysis）,PCNN and other algorithms in terms of articulation grade,spatial frequency and standard deviation,and the algorithm can preserve the details,textures and main characteristics of the source images,and effectively improve the effect of image fusion.