Enriching Visual Expressiveness

Since ancient times, when artists use clay, stone and brush strokes to capture a model or a scene, and even more recently when they use photography and video as their artistic tools, they focus on expressing details which go beyond the physical description of the subject. This includes, for example, the description of the scene ambiance, the nature of the characters and their relations, the motion and positioning in the scene and more. Such attributes are described visually by using a multitude of artistic techniques such as artistic style, visual illusions, scene setup, and even known social rules. Many of the problems encountered in computer graphics deal, in essence, in the description of scenes and models, or the construction of visual transformations between visual media forms. For example, video summarization, Non-photorealistic rendering, video retargeting and image colorization are all examples of such transformations in various levels of expressiveness. This work discusses the topic of visual expressiveness and in particular examines how to automatically extract the visual expressive details in various presentation forms, and construct compositions which reflect some essence of the original by enriching the resulting presentation expressiveness. We present a meta-framework for building expressive medium transformations and present several techniques, which use this framework in the fields of human animation and scene reconstruction. The work consists of four main parts: (1) Expressive Medium Transformations Meta Framework In this chapter we examine some of the common attributes of techniques which deal with expressive transformations between the digital era media forms: photographs, 3D models, video clips, and 3D animations. Following we present some comments and observations about such transformations, and suggest the meta-framework for building such automatic transformations. The main portion of this work, presented in the following three chapters, include examples of such transformations which follow the presented meta-framework for several different media forms, transformations and domains: expressing animation and video in still image, expressing a human animation in a video clip, and constructing an expressive 3D model, given a single image.