一种改进的基于Ripplet变换的多聚焦图像融合方法

针对多聚焦图像在融合时容易产生边缘模糊现象 并损失清晰区域对比度的缺点,基于能有效描述图像边 缘的Ripplet变换(RT),提出一种改进的多聚焦图像融合方法。首先,通过离散RT(type-I) 获 得分解后的高频和低频 子带系数,对于低频子带系数,采用非下采样轮廓波变换(NSCT)进行3层系数融合,而对于 高频子带系数,采用正交匹配 追踪(OMP)算法求解其对应过完备字典的稀疏表示系数；然后通过系数向量的活跃度规则进 行系数融合,并构造其对应的系 数矩阵；最后通过反RT得到最终的融合图像。实验结果表明,本文算法能很好 保持多聚焦图像的边缘轮廓 和清晰部分的对比度,并且在客观评价上,本文算法在互信息、边缘保持度等指标上均具有 一定的优势。

An improved fusion method for multi-focus images based on Ripplet transform

According to the defect that it is easy to produce the edge distortion s and lose clear regional contrast in the fusing process for multi-focus images ,thi s paper proposes a novel multi-focus image fusion algorithm.Firstly,this algori thm gets the decomposed high-frequency and low-frequency sub-band coefficients by the discrete Ripplet transform (type-I).For the low-frequency sub-band coefficients,it adopts the 3-layer non-subsampl ed contourlet transform (NSCT) to conduct coefficients fusion.For the high-freq uency sub-band coefficients,the orthogo nal matching pursuit (OMP) algorithm is used to seek the over-complete dictionary sparse representation coefficients.Then the coeff icient vector activity rules are used to perform the efficient fusion and construct the corresponding coefficients matrix.Finally,t he inverse Ripplet transform is used to get the final fused images.The experiment results show that this algorithm could bette r keep the multi-focus images edge contour and improve the contrast of the clear part,and in the objective evaluation,the algorithm of this paper has a certain advantage on mutual information and the degree of edge keeping.