基于S-PCNN与DDCT相结合的多传感器图像融合  

Multi-sensor image fusion based on S-PCNN and DDCT

作　　者：金鑫[1] 聂仁灿[1] 周冬明[1] 黄晶霞[1] 

机构地区：[1]云南大学信息学院,云南昆明650091

出　　处：《激光与红外》2015年第9期1123-1128,共6页Laser & Infrared

基　　金：国家自然科学基金项目(No.61365001;No.61463052);云南省自然科学基金项目(No.2012FD003)资助

摘　　要：基于简化脉冲耦合神经网络(S-PCNN)与方向离散余弦变换(DDCT)提出一种有效的多传感器图像融合算法。首先将输入图像分为不重叠的方块,并对每个图像块进行8个模的方向DCT变换,得到8方向的模系数;然后分别将图像块对应的模系数送入PCNN模型聚类分析后,对比模系数的点火次数,选取合适图像块系数,得到8个新的图像块系数;最后使用PCA算法将8个图像块合成一幅完整图像块。对输入图像的图像块重复融合过程可得完整的融合图像。An effective multi-sensor image fusion algorithm is proposed based on simplified pulse coupled neural network（S-PCNN） and the direction of the discrete cosine transform （DDCT）. Firstly, the input images are divided into non-overlapping blocks, and the mode coefficients of the eight directions are obtained by proceeding DCT transform of the eight modes＇ direction for each image. And then these mode coefficients that correspond image blocks are input into the S-PCNN model respectively, and the appropriate mode coefficient is selected by comparing the number of ignition of the module coefficients, as a result eight new image block coefficients are gotten. Finally, the eight fused image blocks are compounded in a complete image block by using the PCA algorithm. Repeating the fusion process for each image block of the input image can obtain the complete fusion image.

关 键 词：简化脉冲耦合神经网络 方向离散余弦变换 主成分分析 图像融合 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]