一种半调图像分类识别算法

Classification and recognition algorithm for the halftone image

分析影响逆半调质量的各种因素,引出半调图像识别研究的目标,并运用增强一维自相关函数、共生矩阵和游程矩阵研究半调图像的相关周期和纹理特征,通过构造多级分类器及其标准向量建立了一种常见半调图像的类型识别算法.实验表明,该算法的平均识别正确率可达99%,解决了估值类逆半调技术的实用化问题,也为其他逆半调方案的针对性设计和参数自适应优化奠定了基础.

After analyzing the influences of quality on inverse halftoning, the research on how to classify the halftone has gone into our view. Using the Self-correlation Function of one-dimension, the Grey Level Coocurrence Matrices（GLCM） and the Grey Run-length Matrices（GLRM） the periodic and texture features of the halftoning image are discovered. Based on these properties a new classification algorithm for usual halftones is proposed. Experimental results indicate that the average rate of correct recognition has reached 99%. The new algorithm not only solves the application problem for Estimated-Inverse-Halftoning, but also makes a basic of design and optimization for inverse halftoning.