破损唐卡图像修复质量评价模型的研究

Thangka is the splendid treasure of Tibetan culture, which demonstrates his Tibetan Buddhism history, ethnic, customs, science and technology, and religion to people by impressive visual art. However many treasures had been damaged severely. There is an urgent need to preserve and take actions for protection. For digital protection, in addition to digital repair, it is very necessary to describe quantitatively or evaluation the effect of image inpainting in order to manage, control and improve the quality of the image. In view of repaired thangka image, the project is determined to solve three key problems and make the advanced image quality objective scores available for digital preservation system use. The main contents are as follows:(1)acquire low-levelfeatures and high-level semantic and apply to image quality assessment; (2) establish distortion model and obtain statistical data; (3)study on the method of image quality assessment. Through the research of the above problems, the three models of full- reference, no-reference and weak reference image quality assessment will be proposed. The research results can provide reference or criterions for the effect of Thangka image impainting and gist of choosing algorithm, in the meantime, which also promote the research on image quality evaluation system and contribute to the protection of non-material cultural heritage.

唐卡是藏族文化中灿烂夺目的瑰宝，用震撼人心的视觉艺术给人们展示了藏传佛教历史、民族、民俗、宗教和科技。但众多珍品损毁严重，抢救与整理的任务繁重。对唐卡数字化保护而言，除了采用图像修复技术填补缺损信息以外，如何定量地描述或评价修复的效果是一个亟待解决的问题。本项目以修复后的破损唐卡图像为研究对象，通过三个关键问题：(1)图像底层特征和对象语义获取及应用于评价效果的研究；(2)修复后多种失真的建模和相关统计研究；(3)图像质量评价方法的研究，建立全参考型、“弱参考型”和无参考型质量评价模型，最终实现唐卡修复效果的客观评价。本项目的研究不仅为唐卡数字修复效果提供客观的评价结论，为算法的选择提供依据，扩展和提高计算机对文化遗产的数字化保护研究，而且还将会推动图像质量评价体系以及数字图像处理学科相关问题的研究。

唐卡是藏族文化中灿烂夺目的瑰宝，用震撼人心的视觉艺术给人们展示了藏传佛教历史、民族、民俗、宗教和科技。但众多珍品损毁严重，抢救与整理的任务繁重。对唐卡数字化保护而言，除了采用图像修复技术填补缺损信息以外，如何定量地描述或评价修复的效果是一个亟待解决的问题。本项目从唐卡数字化保护角度，对破损唐卡的图像修复质量评价进行了研究：(1)破损唐卡的自动分割方法；(2) 唐卡图像质量评价数据库的搭建；(3) 唐卡图像及破损图像质量评价模型；(4) 破损修复后质量评价方法和模型。本项目完成了唐卡图像、破损图像、修复后的图像质量客观评价的工作。本项目的研究不仅为唐卡数字修复效果提供客观的评价结论，为算法的选择提供依据，扩展和提高计算机对文化遗产的数字化保护研究，而且还将会推动图像质量评价体系以及数字图像处理学科相关问题的研究。