针对红外图像去噪效果差的问题, 对压缩感知理论中的信号重构问题进行分析, 并提出了一种基于广义 Jaccard系数的多候选集广义正交匹配追踪红外图像去噪算法。本文采用广义 Jaccard系数的相似性匹配准则, 计算感知矩阵和残差的最佳匹配原子, 并将选择的原子作为多个候选集, 迭代时分别将多个原子添加到相应的候选集中, 从多个候选集中选择残差最小的一个作为最终支撑集来重建图像, 对红外图像具有较好的去噪效果。仿真结果表明, 此方法比传统算法收敛速度更快, 峰值信噪比更高, 并且还具有更好的主观去噪效果。

To improve the effect of denoising in infrared images, signal reconstruction in compressed sensing theory is herein analyzed, and a generalized orthogonal matching pursuit algorithm for infrared image denoising based on the generalized Jaccard coefficient is proposed. By employing the similarity matching criterion of generalized Jaccard coefficients, the best matching atom of the perceptual matrix and the residual amount is calculated, and the selected atoms are used as multiple candidate sets. During the iterative process, multiple atoms are added to their corresponding candidate sets, and the one with the smallest residual selected from the multiple candidate sets is utilized as the final support set for image reconstruction, which enables superior denoising of the infrared image. Simulation results show that this method exhibits an enhanced subjective denoising effect, in addition to faster convergence speed, and higher peak signal-to-noise ratio