采用DCT稀疏表示和Dual-PCNN的图像融合算法.pdfVIP

第37 卷 第4 期 红 外 技 术 Vol.37 No.4 2015 年4 月 Infrared Technology Apr. 2015 采用DCT 稀疏表示与Dual-PCNN 的图像融合算法 宋 斌，吴乐华，唐晓杰，文玉强，牟宇飞 （重庆通信学院，重庆 400035 ） 摘要：针对已有图像融合方法会导致融合图像亮度不均匀，与原图像对比度不一致，不适合人眼视觉 效果的缺点，提出将 DCT 稀疏表示与双通脉冲耦合神经网络相结合的压缩感知域图像融合算法。首 先结合图像 DCT 稀疏表示的特点，设计射线采样矩阵；再对测量值采用基于测量值的信息熵加权平 均融合；最后经过全变分优化算法对融合测量值重构得到融合图像。通过对多组不同类型传感器所获 图像融合实验的主观视觉分析和客观评价表明，该算法所得的融合图像能从原始图像中获取更多有用 信息，更好地保持原图像的边缘信息，从而获得更好的视觉效果。 关键词：压缩感知；双通道脉冲耦合神经网络；信息熵；全变分优化算法 中图分类号：TP391 文献标识码：A 文章编号：1001-8891(2015)04-0283-06 An Image Fusion Algorithm Based on DCT Sparse Representation and Dual-PCNN SONG Bin，WU Le-hua ，TANG Xiao-jie，WEN Yu-qiang ，MOU Yu-fei (Chongqing Communication Institute, Chongqing 400035, China) Abstract ：The existing image fusion method results in uneven image brightness, not agreeing with the original image contrast, not suitable for the human eye visual defects. To solve this problem, a new algorithm based on compressive sensing, which combined the DCT sparse representation with the Dual-channel pulse coupled neural network mode, is offered in this paper. First, for the character of the DCT sparse representation, Radial Sampling Matrix is designed. Second, the measurements based on the weighted average is fused with the information entropy of measurements. Finally, the total variation algorithm is used tore construct the fusion image. Experiments have been done to fuse multiple sets of different types of sensor image. Both subjective visual analysis and objective evaluation criteria show that the proposed algorithm ca