Am Nat. 2019 Feb;193(2):164-186. doi: 10.1086/701300. Epub 2019 Jan 16.
Animal coloration patterns, from zebra stripes to bird egg speckles, are remarkably varied. With research on the perception, function, and evolution of animal patterns growing rapidly, we require a convenient framework for quantifying their diversity, particularly in the contexts of camouflage, mimicry, mate choice, and individual recognition. Ideally, patterns should be defined by their locations in a low-dimensional pattern space that represents their appearance to their natural receivers, much as color is represented by color spaces. This synthesis explores the extent to which animal patterns, like colors, can be described by a few perceptual dimensions in a pattern space. We begin by reviewing biological spatial vision, focusing on early stages during which neurons act as spatial filters or detect simple features such as edges. We show how two methods from computational vision-spatial filtering and feature detection-offer qualitatively distinct measures of animal coloration patterns. Spatial filters provide a measure of the image statistics, captured by the spatial frequency power spectrum. Image statistics give a robust but incomplete representation of the appearance of patterns, whereas feature detectors are essential for sensing and recognizing physical objects, such as distinctive markings and animal bodies. Finally, we discuss how pattern space analyses can lead to new insights into signal design and macroevolution of animal phenotypes. Overall, pattern spaces open up new possibilities for exploring how receiver vision may shape the evolution of animal pattern signals.
动物的颜色图案，从斑马的条纹到鸟蛋上的斑点，都非常多样化。随着对动物图案的感知、功能和进化的研究迅速发展，我们需要一个方便的框架来量化它们的多样性，特别是在伪装、拟态、配偶选择和个体识别的背景下。理想情况下，图案应该通过它们在低维图案空间中的位置来定义，这个空间代表了它们在自然接收者眼中的外观，就像颜色用颜色空间来表示一样。本综述探讨了动物图案在多大程度上可以用图案空间中的几个感知维度来描述，就像颜色一样。我们首先回顾了生物空间视觉，重点介绍了神经元作为空间滤波器或检测简单特征（如边缘）的早期阶段。我们展示了计算视觉中的两种方法——空间滤波和特征检测——如何提供动物颜色图案的定性不同的度量。空间滤波器提供了图像统计的度量，这些统计量由空间频率功率谱捕获。图像统计提供了图案外观的稳健但不完整的表示，而特征检测器对于感知和识别物理对象（如独特的标记和动物身体）是必不可少的。最后，我们讨论了如何通过图案空间分析来深入了解动物表型信号设计和宏观进化。总的来说，图案空间为探索接收者视觉如何塑造动物图案信号的进化开辟了新的可能性。
