改进的正交匹配追踪图像快速重建算法  

Image fast reconstruction algorithm based on improved orthogonal matching pursuit

作　　者：黄会营[1] 庄淑君[1] 

机构地区：[1]洛阳理工学院电气工程与自动化系,河南洛阳471003

出　　处：《光学技术》2014年第6期515-519,共5页Optical Technique

摘　　要：传统采样需要的采样间隔小,当信号过长时会增加采样量,消耗过多的时间,占用更多的存储空间。提出了以Daubechies小波变换基作为稀疏变换基,用高斯随机矩阵作为观测矩阵,采用基于Dice匹配准则的正交匹配追踪算法作为恢复算法对图像进行重建。在分析了正交匹配追踪算法的特点后,提出了一种基于Dice匹配准则的改进的正交匹配追踪算法。该算法在原子匹配过程中将几何平均值用算术平均值替代,使重要成份在向量中的作用更加突出,能更准确地从字典中挑选出与残差向量最匹配的原子。仿真结果表明,该算法在相同的观测条件下,能减少重构误差,可将图像的重构质量提高2%左右,重构概率提高7%左右,有较高的应用价值。Sampling interval required for traditional sampling is small,when the signal is too long,the sample volume is increased,the more time is consumed,the more storage space is taken up.A improved orthogonal matching pursuit（OMP）algorithm based on atomic matching criterion of Dice coefficient（DOMP）is used to reconstruct images based on Daubechies wavelet transform base as sparse transform base and Gaussian random matrices as measurement matrices.In the atomic matching of DOMP algorithm the arithmetic mean is introduced to replace the geometric mean,the role of an important component in the vector is more prominent using of Dice measurement,so it enables more accurately to pick out the best match with the residual vector atoms from the dictionary.The experiment results show that the reconstruction error of images has great improvements under the same number of measurement values.The quality of the reconstructive image is improved about 2%,the probability of the reconstructive image is improved about 7%,this algorithm has higher application value.

关 键 词：压缩感知 正交匹配追踪 Dice系数 图像重建 小波变换 

分 类 号：TN911.73[电子电信—通信与信息系统]