基于TV范数的压缩感知光偏折层析重建

Compressed Sensing Optical Deflection Tomography Reconstruction Technology Based on TV Norm

针对光偏折层析的稀疏投影重建问题,本文将偏折角转化的迭代算法与TV范数的压缩感知相结合重建被测场.该算法使用被测量场的全变差作为稀疏度的先验模型,并结合最速下降法来调整全变差,稍微调整梯度下降法的步长,并用一致性控制步长以使梯度下降.在8个方向投影的条件下,模拟了三峰高斯温度场,用不同步长的算法进行了重建,并在同一条件下对不同的算法进行了比较.实验结果表明,一致性控制步长算法的重建质量更好,可以更有效地抑制噪声.

Aimed at the problem of sparse projection reconstruction in optical deflection tomography,the iterative algorithm of deflection angle transformation was combined with the compressed sensing of TV norm to reconstruct the measured field.The algorithm used the total variation of the measured field as aprior model of sparsity,and combined the steepest descent method to adjust the total variation,the step size of the gradient descent method was adjusted slightly,and decreased the gradient by adopting the consistency controlled steepest descent(CCSD).The three-peak Gauss temperature field was simulated under the condition of eight directional projections,the reconstruction was carried out by using the unsynchronized algorithms,and different algorithms were compared under the same condition.From the experimental results,it can be seen that the reconstruction quality of the consistency controlled steepest descent(CCSD)algorithm is better and the noise can be suppressed more effectively.