[论文解读] The Artists who Forged Themselves: Detecting Creativity in Art
本文提出一种新颖方法,通过基于小波的统计复杂性建模纹理特征中的自组织性,以检测绘画中的艺术创造力。通过分析小波系数中的因果依赖关系,该方法将原创作品识别为无需事先知晓原作的自组织过程的实现,在马格里特作品中不同颜色区域的检测结果具有一致性。
Creativity and the understanding of cognitive processes involved in the creative process are relevant to all of human activities. Comprehension of creativity in the arts is of special interest due to the involvement of many scientific and non scientific disciplines. Using digital representation of paintings, we show that creative process in painting art may be objectively recognized within the mathematical framework of self organization, a process characteristic of nonlinear dynamic systems and occurring in natural and social sciences. Unlike the artist identification process or the recognition of forgery, which presupposes the knowledge of the original work, our method requires no prior knowledge on the originality of the work of art. The original paintings are recognized as realizations of the creative process which, in general, is shown to correspond to self-organization of texture features which determine the aesthetic complexity of the painting. The method consists of the wavelet based statistical digital image processing and the measure of statistical complexity which represents the minimal (average) information necessary for optimal prediction. The statistical complexity is based on the properly defined causal states with optimal predictive properties. Two different time concepts related to the works of art are introduced: the internal time and the artistic time. The internal time of the artwork is determined by the span of causal dependencies between wavelet coefficients while the artistic time refers to the internal time during which complexity increases where complexity refers to compositional, aesthetic and structural arrangement of texture features. The method is illustrated by recognizing the original paintings from the copies made by the artists themselves, including the works of the famous surrealist painter René Magritte.
研究动机与目标
- 客观检测绘画中的艺术创造力作为自组织过程,独立于艺术家归属或原作知识。
- 建立一个通过测量数字图像表示中统计复杂性的框架,以识别原创创造性作品。
- 引入两个时间概念——内在时间与艺术时间——分别与因果依赖关系和复杂性增加相关联。
- 证明原创作品在不同颜色区域表现出一致的自组织模式,即使由同一艺术家复制亦如此。
- 提供一种无需训练集、笔触模型或距离度量的方法,实现可泛化的创造力检测方法。
提出的方法
- 使用小波分解(如Haar、Daubechies、Coiflet)在数字绘画的多个尺度上分析纹理特征。
- 应用基于因果态的统计复杂性度量,利用最优预测特性来量化自组织性。
- 将内在时间定义为小波系数之间因果依赖的跨度,反映图像中的时间结构。
- 引入艺术时间为复杂性增加的持续时间,对应于构图与美学组织。
- 在不同颜色通道(红、绿、蓝)的图像块(如512×512像素)上计算全局复杂性,以评估一致性。
- 通过复杂性作为创造力的代理指标,比较原作与自创复制品,检测自组织模式的差异。
实验结果
研究问题
- RQ1艺术创造力能否在不事先知晓原作的情况下,客观地作为数字绘画中自组织过程被检测?
- RQ2通过因果依赖关系和复杂性增加定义的内在时间与艺术时间,如何与绘画的美学与结构组织相关联?
- RQ3纹理特征的统计复杂性在同幅画的不同颜色区域中一致性如何?
- RQ4自组织指数能否在高度精确的复制品中区分原作,即使视觉差异微小?
- RQ5该方法在不同小波族(如Haar、db2、sym3)和颜色通道中是否保持稳健?
主要发现
- 基于统计复杂性的自组织指数在马格里特绘画的512×512像素图像块中表现出显著一致性,无论主导颜色为蓝色、绿色或红色。
- 对于画作《眼泪的滋味1号与2号》,蓝色和绿色图像块的全局复杂性值在多种小波族中保持稳定,数值范围在0.188至0.323之间。
- 该方法能成功区分原作与自创复制品,即使视觉差异极小,通过检测自组织模式的偏差实现。
- 该方法无需训练集、复杂的小波变换或针对纹理与笔触的独立模型,简化了实现过程。
- 结果在不同颜色块中的一致性表明,马格里特在复制自身作品时可能采用了刻意的自组织技术,与超现实主义实践一致。
- 该框架可在不知晓原作性的情况下实现创造力检测,适用于未知或存在争议的艺术品。
更好的研究,从现在开始
从阅读论文到最终审阅,大幅缩短您的研究时间。
无需绑定信用卡
本解读由 AI 生成,并经人工编辑审核。