基于模糊神经网络的印刷线条感知质量评价

Printed line perceptual quality assessment based on fussy neural network

线条质量作为印刷品质量评价的主要内容之一,长期以来采用视觉主观评价方法,具有主观性、不一致性、效率低等问题。数字印刷领域虽然已开始采用机器视觉技术开展印刷线条质量的客观指标检测,但还不能明确给出这些客观测量值与线条整体感知质量之间的关系描述。对多组具有不同的模糊度和粗糙度边缘特征、不同的暗度和对比度表现、不同的线宽复制误差和填充效果的印刷线条开展实验研究,通过对线条质量属性的客观测量值与线条主观感知质量评价分值之间的数据分析,利用模糊神经网络算法建立了线条质量属性与主观感知评价结果之间的关系模型。实验结果表明,印刷线条感知质量的评价建模应以ISO13660标准提出的6个线条质量属性为依据,任何一个属性都不能唯一地确定印刷线条的质量,同时也都不应该从综合评价中删除出去;所提出的印刷线条感知质量评价建模方法是有效可行的,评价模型的预测结果与主观评价结果间具有很好的一致性,为印刷线条质量属性的客观检测赋予现实意义,也为基于机器视觉的印刷线条感知质量的综合评价提供了理论依据。

Line quality as one of the main contents of the print quality assessment is evaluated with visual subjective as- sessment method for a long time, which has the problems, such as subjectivity, inconsistency and low efficiency. Al- though digital print field has started using machine vision technique to assess the printed line quality with its objective indices ,it is still difficult to describe the relationship between these objective measurement values and the overall prin- ted line perceptual quality. This paper experimentally studies various groups of printed lines with different blurriness and raggedness of line edges, different darkness and contrast, different line width duplicating error and filling effect. Through analyzing the data relationship between the subjective perceptual quality assessment score and the six objective measurement values of the line quality attributes of the printed lines from a large group of prints, a fussy neural network based assessment method of printed line perceptual quality is proposed,with which the relationship model between the line quality attributes and subjective perceptual assessment result is established. Experimental results show that the printed line perceptual quality assessment modeling should be based on the six line quality attributes proposed by ISO 13660;and the line perceptual quality can not be decided only by any one of the attributes;on the other hand,any one of the line attrib- utes should not be excluded from the assessment model. The proposed modeling method is effective and feasible ;the objec- tive assessment results predicted from the line quality assessment model have good consistency with the subjective assess- ment results,the result indicates that the printed line perceptual quality can be assessed by machine vision instead of hu- man viewers ,which meets the objective measurement and assessment requirement in printing industry.