模式识别与LUT相结合的误差分散类逆半调方法研究(英文)

Inverse Halftoning for Error Diffusion Based on Pattern Recognition and Look-up Table

查找表(LUT)图像逆半调是一种基于模式识别原理的快速非迭代方法。其中,产生 LUT数据的模板设计、LUT数据训练图集的选择,及其训练后“未出现模式逆半调值”的估计方法成为决定逆半调图像质量的关键。在分析图像半调模型的基础上,应用LUT逆半调算法对LUT数据训练图集进行了充分的实验,统计和分析了训练图集对 LUT逆半调图像质量的影响,给出了其选择依据。实验表明依据该准则训练出的LUT数据在多项逆半调效果评价上表现良好。

Look-up table (LUT) inverse halftoning is a fast method based on pattern recognition without iteration. The LUT is obtained from the histogram gathered from a set of samples including halftone images and corresponding original images. Its efficiency depends heavily on the choice of template for building LUT, the choice of image training set and the choice of method for estimating continuous-tones of nonexistent patterns. The LUTs are derived by various image training sets and are applied to inverse halftoning. By statistically analyzing the influence of image training sets on the quality of LUT inverse halftoning a guideline is proposed on selection of image training sets. Experiments show that the LUT inverse halftoning performs better according to the guideline on building the LUT.