Digital Morphology of Ornaments – Novel Methods for the Analysis, Modelling and Research of Ornament Forms, linking Art History and Computer Vision, applied to Augsburg Rocaille Prints of the 18th Century

很少有艺术历史研究领域像对花饰的研究那样具有如此基本的需求。这些形式在1730年左右出现在法国，并成为18世纪的主要装饰形式，特别是在中欧。尽管有大量的文物——从鼻烟壶到巨大的灰泥作品——以及无穷无尽的各种形式，但在年代、归属和科学评估方面仍然存在令人遗憾的不确定性。这种缺陷的主要原因是术语上的困难-不规则的，贝壳状的形式逃避了几何和图像的描述。然而，法国形式的装饰版画被传播和解释，为全面分析提供了很好的机会。仅在重要的出版中心奥格斯堡就有几千幅蚀刻版画，展示了1730年至1775年间印刷的工艺品和装饰形式。我们通过在艺术史和计算机科学之间的接口采用跨学科的方法来解决这一需求：用计算机视觉领域的技术分析形式，特别是结合图论和机器学习。为了可行地应用于可调形式，我们将开发新的方法，以使抽象图像内容可以通过算法技术进行调查。基于现有的约2100件数字化奥格斯堡版画，我们将开发一个工具链，根据相似性对作品进行分类，并使相关形式和作品的识别成为可能。这个研究工具将使我们能够找到关于版画作为模型和图案的归属，日期和应用的基本问题的答案，这反过来又开辟了当前艺术史理论的研究视角，特别是在文化转移和绘画转向领域。我们的实现将基于将叶栅组合物细分为蜗壳和附加的多晶脊的框架的可能性。我们将提取描述和抽象蜗壳S形和c形集合的几何、属性图，同时我们将用统计噪声参数化和子图将变形脊表示为纹理。结合合适的相似性度量、图形可视化范例和图形转换，我们将开发工具，为艺术史上的饰品研究提供新的可能性。

Few areas of art historical research are characterized by such elementary desiderata as studies on rocaille ornaments. These forms emerged in France around 1730 and became the predominant form of ornamentation of the 18th century, especially in Central Europe. Notwithstanding the large number of objects – from the snuffbox to the monumental stucco work – and a sheer inexhaustible variety of forms, a deplorable uncertainty in dating, attribution and scientific assessment still prevails. The main reason for this deficiency are terminological difficulties – the irregular, shell-like forms evade both a geometric and an iconographic description. Yet, ornamental prints in which the French forms were disseminated and interpreted offer great opportunities for a comprehensive analysis. Several thousand etchings originate from the important publishing center Augsburg alone, showing arts and crafts objects and decorative forms, printed between 1730 and 1775.We address this desideratum by employing an interdisciplinary approach at the interface between art history and computer science: the analysis of forms with techniques from the field of computer vision, especially in combination with graph theory and machine learning. For a feasible application to rocaille forms, we will develop new methods in order to make the abstract image contents accessible to an investigation with algorithmic techniques. Based on an already existing collection of about 2100 digitized Augsburg prints, we will develop a tool chain which categorizes the works according to similarities and enables the identification of related forms and works. This research tool will enable us to find answers to fundamental questions concerning the attribution, dating and application of prints as models and patterns, which in turn open up research perspectives in current art historical theory, especially in the areas of cultural transfer and the pictorial turn.Our realization will be based on the possibility of subdividing the rocaille compositions into a framework of volutes and the attached polymorphic ridges. We will extract geometric, attributed graphs describing and abstracting the set of S- and C-shapes of the volutes, while we will represent the protean ridges as textures with statistical noise parametrizations and sub-graphs. Together with suitable similarity measures, graph visualization paradigms, and graph transformations, we will develop tools that afford new possibilities for researching ornaments in art history.