Non-photorealistic Shading and Hatching

Abstract

Seeing the external world is an essential ability of the human being. It ensures the most vital survival functions such as avoiding predators, search of food or reproduction. Our visual system is a fine result of millions of years of evolution and constant improvement. It is not surprising that it is extremely fast and provides to our brain the information about spatial structure of the external world, as well as the information about relative movement of external objects and the observer himself. The science of vision (neuroscience, psychology, computer vision) studies the process of perception of the external world, starting from retinal stimuli and ending up in a complete spatiotemporal internal brain image. According to the most influential Marr computational model of vision, the whole process is divided into several stages: from unstructured retinal image to primal 2.5 sketch, then to object-based categorial model, essential to develop an understanding of the scene. This process is extremely complex, with several feedback loops [19,26]. To facilitate the task at early stages of the perception process, our vision system uses, as visual cues, various pieces of information available in the scene: variations of tone due to natural distribution of light in the scene, relative size and foreshortening of objects due to the perspective, presence of edges due to particular viewing point, etc. It has been discovered that some of these perception functions are 'wired' (e.g., lateral inhibition in the retinal processing, which facilitates edge detection); others are high-level functions of our brain, which are learned from experience [19,26].Artists were always interested in depicting the external world. Early examples of such depictions can be found in Lascaux caves; they are estimated to be 17,300 years V. Ostromoukhov ( ) CNRS/Université Claude Bernard Lyon 1, 43, bd du 11 novembre 1918,

Generating knitting patterns from a sketch

“…Moreover, as the space of structural elements constituting an image shrinks, the potential weight of symbolic meaning conveyed by the remaining elements increases [Arnheim 1974, p.47-95]. A common way to handle semantic artifacts is to choose different hatching algorithms for geometrically structured and unstructured sources [Ostromoukhov 2013]. Prominent edge structure in images benefits from using hatching algorithms that build on it, while stochastic algorithms are preferred in blurry images [Jodoin et al 2002].…”