高级检索

风格一致的书法匾额生成

Generating Style Consistent Calligraphy Tablet

  • 摘要: 针对匾额上的书法字通常希望书法名家来题写而找名家不易的问题,提出一种使用优秀历史书法作品图像生成风格一致的书法匾额的方法:首先,从书法库中检索匾额所需内容,提取书法图像的笔画特征、结体特征和章法特征作为度量依据;接着,以内容中第1个书法字的风格为基准,计算多个书法词组内的风格一致性程度,按一致程度值排序;然后,对选中的书法词组图像进行分段连续光滑矢量化,以使缩放不失真,曲线段间保持C^1连续.最后,生成风格一致、颜色一致、大小一致的书法匾额.用CADAL书法库中3146幅单字图像进行匾额合成实验,结果表明,该方法对风格一致性分类的平均准确率为0.97,用户对所生成的匾额基本满意.

     

    Abstract: Calligraphic characters on tablets are often written by famous calligraphers to achieve celebrity effect,but it is not convenient to obtain famous calligraphy characters.This paper proposes an approach to generate style consistent calligraphy tablets by using character images from famous historic calligraphy works.Firstly,candidates of calligraphy character for the tablet are retrieved and grouped,followed by feature extraction of strokes,characters and layouts for style consistency computation within the group and aligned with the first character.Then,most style consistency groups are ranked for user feedback.After that,smoothness vectorization algorithm is proposed to avoid the distortion in scaling and maintain C^1 continuity of the connecting points between curve segments.Finally,style consistency,color consistency and size consistency of calligraphy character group is generated for the user specified tablet.The proposed algorithm is validated on 3146 character images from CADAL calligraphy database.The average group style consistency accuracy is 0.97 with satisfaction from tested users.

     

/

返回文章
返回