结合小波域马尔可夫树模型的压缩采样图像重建

摘要

Model-based compressive sampling reconstruction can achieve better reconstruction results.A novel image reconstruction method using wavelet domain Markov tree model is proposed for compressive sampling.Markov tree model matches the statistical property of wavelet coefficients well.Therefore,the model can contribute to the atom selection in the OMP algorithm framework.In the proposed algorithm,new atoms are selected sequentially from the candidate atoms that match the residual signal best.The atoms are selected using the maxim likelihood principle.Experiments show that the new method can find atoms with large coefficients more accurately; therefore the recovered images are better than traditional algorithms.%已有的研究表明基于模型的压缩采样信号重建可以取得更好的重建效果.本文提出一种结合小波域马尔可夫树模型的压缩采样图像重建方法.马尔可夫树模型很好的匹配了图像小波变换后的系数在尺度间的持续性.这种统计特性可以在正交匹配追踪算法中协助原子的选取,从而更准确的选取具有大幅值系数的原子.在本文提出的新算法中,每次迭代新增的原子是从与残差信号较匹配的候选原子中选取.候选原子中使模型的状态似然函数最大的原子被选出.实验结果表明,新算法可以更准确选出具有大系数原子,重建的图像质量好于其它传统方法.