基于CRF预分割和NSCT的多聚焦图像融合

针对基于多尺度分解的图像融合算法在提高图像模糊区域质量的同时却降低了清晰区域质量的问题,提出了一种结合CRF预分割与NSCT的多聚焦图像融合算法.该算法首先利用条件随机场概率模型对图像进行预分割,将图像预分割为清晰区域、模糊区域和两者的交界区域;然后对图像进行NSCT分解,低频子带系数采用加稀疏约束的非负矩阵分解算法融合,高频子带系数,在交界区域采用方向特征对比度的方法融合,在非交界区域选取清晰度高的系数融合.最后经NSCT逆变换得到最终的融合图像.该方法的融合效果在提高了图像清晰度的同时又有效保持了图像的边缘信息.

Multi-focus image fusion based on CRF pre-segmentation and NSCT

In this paper,focusing on the issue based on multi-scale decomposition of the image fusion algorithm to improve the quality of the image blurring area while reducing the quality of a clear area,a multi-focus image fusion algorithm combining CRF pre-segmentation and NSCT transformation was proposed.Firstly,the image is pre-segmented into three regions:the clear region,the fuzzy region and the boundary region which based on the probability model of conditional random field.And then,the image is decomposed by NSCT transformation,the fusion principle of low frequency sub-band coefficients is based on a constrained sparse algorithm for nonnegative matrix factorization algorithm,the high frequency sub-band coefficients are fused in the boundary region using a method which selected the maximal contrast values of its directional feature,and the coefficients with high sharpness in the two images are directly selected as the fusion result in the non-boundary regions.The fusion image was reconstructed using the inverse NSCT.The experimental results show that the fusion effect of the method can improve the image clarity while maintaining the edge information of the image.