为改进查找表逆半调算法中“未出现半调模式逆半调值”的估计精度，提出了一种查找表与Elman回归网络相结合的图像逆半调算法。该算法首先通过样本图集生成初步逆半调查找表，然后以Elman型回归网络为工具，构造、训练逆半调逼近模型，最后达到拟合“未出现半调模式逆半调值”的目的，产生完整查找表，支持逆半调处理。实验结果及性能分析表明，应用本文算法生成的逆半调重建图像在视觉效果及PSNR指标上表现良好，具有运行速度快、空间复杂度低的特点。

In order to improve the estimation precision of inverse halftoning values whose halftone patterns don’t appear in image training set, an inverse halftoning method based on look up table(LUT) and Elman recurrent network is proposed. Firstly the preliminary LUT is built from image training set. Then the Elman recurrent network is used to approximate the mapping between halftoning and inverse halftoning. After studying, training and optimizing the structure, layers and nodes of the Elman network are established appropriately. With this network the continuous-tones of nonexistent halftone patterns are estimated. And the complete LUT is generated. Experiments show that the proposed algorithm can generate inverse halftoning images of good in both visual effect and PSNR while having low memory requirement and inexpensive computation than some of existing algorithms.