Shapes illustration with feature lines and stylized rendering.
Date of Issue2008
School of Computer Engineering
Over centuries, artists and illustrators have developed powerful techniques to convey shapes effectively in very rich forms. In recent years, non-photorealistic rendering has been improved greatly to assist artists and illustrators in their work with computer generated illustrations. However, it is more than often that hand-drawn illustrations are still the only choice in modern industry. In this dissertation, we aim to provide more effective techniques for shape illustration with feature lines and stylized rendering. The first part of our work is to obtain new feature lines which are more perceptually consistent and flexible in conveying shapes. A number of feature lines have been proposed, but they are not really able to capture all the visually relevant features by themselves. We propose Photic Extremum Line (PEL) which emphasizes significant variations of illuminance over 3D surfaces based on observation in human vision and perception that a sudden change in luminance plays a critical role in faithfully representing and recovering 3D information. Compared with the existing feature lines, PEL is more flexible and offers users more freedom and control to achieve their desired illustrations. Two efficient PELs extraction algorithms are presented to incorporate PELs to various rendering systems. Next, we simulate the traditional sketching based on PELs and contours. Sketching is one of the most widely used techniques, and in practice, architects even trace over computer generated line drawings of their initial designs to create a sketchier look. We propose a novel approach to emulate sketching strokes with good texture and break long strokes into short ones. Our approach incorporates principles of perceptual grouping with a global segmentation algorithm. By applying this approach at a succession of scales, the successive approximation effect seen in sketching can also be effectively simulated. For many applications, one of the most important goals of illustration is to achieve visually pleasing line drawings, such as cartoon. We hence present an effective and flexible framework to convey shapes with stylized line drawings based on a novel stroke representation model, disk B-spline curve(DBSC). The flexibility of the stylized line drawing framework is further demonstrated with a wide range of styles. Finally, stylized shading is equally important in illustration, which gives strong cues of shapes and more details. We experiment in stylized shading by modeling shading styles with low level statistical features, and present an improved feature guided texture synthesis algorithm to generate stylized shading from input examples.
DRNTU::Engineering::Computer science and engineering::Computing methodologies::Computer graphics