Images from my thesis


Figure 1.1: An few examples of a NURBS-based model displayed in wireframe. The large number of isolines makes distinguishing key features difficult.


Hand-tuned Phong-rendered image of mechanical part by Dr. Sam Drake. It took him approximately six hours to hand tune this image.	Illustration of a mechanical part using lines to separate parts and shading to convey material properties. Courtesy Macmillan Reference USA, a Simon \& Schuster Macmillan Company~\cite{macm95}.
Figure 1.2: Hand-tuned Phong-shaded image versus technical illustration.


Copyright 1996 Barbara Meier~\cite{meie96}. Used by permission.	Copyright 1990 Paul Haeberli~\cite{haeb90}. Used by permission.	Copyright 1997 Cassidy Curtis~\cite{curt97}. Used by permission.
Figure 2.1. Non-photorealistic one-shot images with a high level of abstraction.

Figure 2.2: Computer-generated pen-and-ink illustrations.



Pen-and-Ink Illustration. Copyright 1996 Michael Salisbury~\cite{sali96}}. Used by permission.	Pen-and-Ink Illustration. Copyright 1994 Georges Winkenbach~\cite{wink94}}. Used by permission.


Figure 2.3: One-shot computer-generated pen-and-ink images conveying shape by Winkenbach. Copyright 1996 Georges Winkenbach [39]. Used by permission.


Copyright 1990 Gershon Elber~\cite{elbe90}. Used by permission.	Copyright 1990 Debra Dooley~\cite{dool90}. Used by permission.
Figure 2.5: One-shot images conveying shape by Dooley and Elber.

Figure 2.6: Image from a frame of Markosian et al. [24] real-time 3D interactive system for illustrating non-self-intersecting polygon mesh-based models. Copyright 1997 Lee Markosian. Used by permission.



Figure 3.1: Technical illustration conventions. Copyright 1995 Volvo Car UK Limited [28].


Figure 3.2: An example of the lines an illustrator would use to convey the shape of this airplane foot pedal. Copyright 1989 Macdonald & Co. (Publishers) Ltd. [25].


Figure 3.3: Changing which isolines are displayed can change the perception of the surface. The image on the right looks as if it has a deeper pit because the isolines go thru the maximum curvature point on the surface. Images courtesy of David Johnson.

Figure 3.4: Illustrators use lines to separate parts of objects and define important shape characteristics. This set of lines can be imitated for NURBS models by drawing silhouettes, boundaries, and discontinuities, shown above (drawn over the wireframe representation).

Figure 3.5: Definition of a silhouette: At a point on a surface, and given E(u,v) as the eye vector and n(u,v) as the surface normal, a silhouette point is defined as the point on the surface where = 0 or the angle between E(u,v) and n(u,v) is 90 degrees.

Figure 3.6: Three line conventions suggested by Judy Martin [25]. Left: single line weight used throughout the image. Middle: heavy line weight used for out edges and parts with open space between them. Right: vary line weight to emphasize perspective. Copyright 1989 Macdonald & Co. (Publishers) Ltd. [25].

Figure 3.7: This photograph of a metal object shows the anisotropic reflections and the white edge highlights which illustrators sometimes depict.

Figure 3.8: Illustrators sometimes use the convention of white interior edge lines to produce a highlight [25].

Figure 3.9: Illustrators combine edge lines with a specific type of shading. Shading in technical illustration brings out subtle shape attributes and provides information about material properties. Left: Compare this shaded image of airplane pedal to the line drawing in Figure 3.2. Copyright 1989 Macdonald & Co. (Publishers) Ltd. [25]. Right: Engine. Courtesy Macmillan Reference USA, a Simon & Schuster Macmillan Company [28].

Figure 4.1: Interpolating silhouettes: After two neighboring surface points with different 's are found, the point where E(u,v) . n(u,v) = = 0 can be found by linearly interpolating in u or v with the two angles as calculated in Equation 4.1. Note: ₁ = (E₁ . n₁) is greater than 0 and ₁ = (E₂ . n₂) is less than 0 .

Figure 4.2: Calculating the mesh normals: The four mesh normals which correspond to m_i,j are n_1,3, n_1,2, n_4,3, n_4,2, where for example n_1,3 = vec₁ x vec₂, with vec1 = m_i,j - m_i-1,j and vec2 = m_i,j-1 - m_i,j.

Figure 4.3: Envision the mesh method as a table of signs, where

can be +, -, or 0.

Mesh Method

Srf-Node Method

Figure 4.4: These images show the control mesh (in uv-space) for a surface where +, -, or 0 denotes the sign of the dot product (E(u,v) . n(u,v)). For the Mesh Method, there are up to four dot products that need to be calculated per mesh point, only one per mesh point for Srf-Node Method.

Mesh Method Srf-Node Method
Figure 4.5: These images show the control mesh (in uv-space) for a surface, with approximations to the silhouettes. The sign of the dot product (E(u,v) . n(u,v)) are denoted by +, -, or 0.


View of surface with silhouettes generated with mesh method.	Looking down on surface with silhouettes.
Figure 4.6: Mesh Method.


View of surface with approximate silhouettes generated with the tessellated mesh method.	Looking down on surface with approximate silhouettes.
Figure 4.7: Tessellated Mesh Method.

Figure 4.8: Visualize the Srf-Node Method as a table of signs

, where can be +, -, or 0.

Figure 4.9: Srf-Node Method can result in missed silhouettes depending upon the node points. If for example, the node points were those that correspond to theta_1, theta_2,and theta_3, there would be three missed silhouette points because theta_1, theta_2,and theta_3, are all less than 90 degrees and there would be no sign change. However, if the nodes points were alpha, theta_2,and theta_3, then alpha is greater than 90 degrees and theta_2 is less than 90 degrees, so the silhouette between the two corresponding node points would not be missed and could be interpolated. The problem of missing these silhouettes can be remedied by refining the control mesh.


View of surface with silhouettes and surface boundaries generated with Srf-Node Method.	Looking down on surface with silhouettes and surface boundaries.
Figure 4.10: Srf-Node Method.

Figure 4.11: Looking down on surface with silhouette generated with Srf-Node method. Compare this image with the 2D projection and approximation of silhouettes shown in Figure 4.5 using the Mesh method and the Srf-Node method.

Figure 4.12: View of the same surface represented in Figure 4.5, 4.4, and 4.11 with silhouettes generated with the Srf-Node method.

Figure 4.13: Diffuse shaded image using Equation 4.1 with k_d = 1 and k_a = 0. Black shaded regions hide details, especially in the small claws; edge lines could not be seen if added. Highlights and fine details are lost in the white shaded regions.

Figure 4.14: Image with only highlights and edges. The edge lines provide divisions between object pieces and the highlights convey the direction of the light. Some shape information is lost, especially in the regions of high curvature of the object pieces. However, these highlights and edges could not be added to Figure 4.13 because the highlights would be invisible in the light regions and the silhouettes would be invisible in the dark regions.

Figure 4.15: Phong-shaded image with edge lines and k_d = 0.5 and k_a = 0.1. Like Figure 4.13, details are lost in the dark gray regions, especially in the small claws, where they are colored the constant shade of k_d k_a regardless of surface orientation. However, edge lines and highlights provide shape information that was gained in Figure 4.14, but could not be added to Figure 4.13.

Figure 4.16: Approximately constant luminance tone rendering. Edge lines and highlights are clearly noticeable. Unlike Figures 4.13 and 4.15 some details in shaded regions, like the small claws, are visible. The lack of luminance shift makes these changes subtle.

Figure 4.17:How the tone is created for a pure red object by summing a blue-to-yellow and a dark-red-to-red tone.

Figure 4.18: Luminance/hue tone rendering. This image combines the luminance shift of Figure 4.13 and the hue shift of Figure 4.16. Edge lines, highlights, fine details in the dark shaded regions such as the small claws, as well as details in the high luminance regions are all visible. In addition, shape details are apparent unlike Figure 4.14 where the object appears flat. In this figure, the variables of Equation 4.1 and Equation 4.1 are: b = 0.4, y = 0.4, alpha = 0.2 and beta = 0.6.

Figure 4.19: Luminance/hue tone rendering, similar to Figure 4.18 except b = 0.55, y = 0.3, alpha = 0.25, and beta = 0.5. The different values of b and y determine the strength of the overall temperature shift, where as and determine the prominence of the object color, and the strength of the luminance shift.

Figure 4.20: Comparing shaded, colored spheres. Top: Colored Phong-shaded spheres with edge lines and highlights. Bottom: Colored spheres shaded with hue and luminance shift, including edge lines and highlights. Note: In the first Phong-shaded sphere (violet), the edge lines disappear, but are visible in the corresponding hue and luminance shaded violet sphere. In the last Phong-shaded sphere (white), the highlight vanishes, but is noticed in the corresponding hue and luminance shaded white sphere below it. The spheres in the second row also retain their ``color name.''

Figure 4.21: Tone and undertone shaded spheres with backgrounds getting darker.

Figure 4.22: Shaded spheres without edgelines. Top: Colored Phong-shaded spheres without edge lines. Bottom: Colored spheres shaded with hue and luminance shift, without edge lines.

Figure 4.23: An anisotropic reflection can be seen in the metal objects in this photograph.


Phong-shaded object.	New metal-shaded object without edge lines.

New metal-shaded object with edge lines.	Metal-shaded object with a cool-to-warm shift.
Figure 4.24: Representing metallic material properties.

Figure 4.25: Comparing this figure to Figure 1.1, the edge lines displayed provide shape information without cluttering the screen.


Frame of model with new shading in an interactive environment with lights positioned up and to the right.	Frame after the camera position is moved to view the side of the model.	Frame after moving the object instead of the camera, allowing the surface to vary completely from cool to warm.
Figure: Frames from the NPR JOT Program, to which I used Markosian et al.'s silhouette finding technique [24] and added the OpenGL approximation to the new shading model. This will be discussed further in Chapter 5.


Image of with edge lines only.	Image with new shading and edge lines.
Figure 5.1: An Alpha_1 model that was tessellated and imported into the JOT NPR Program.


Phong shading model for colored object.	New shading model without edge lines.

New shading model: edge lines, highlights, and cool-to-warm hue shift.	Approximation: Phong shading, two colored lights, and edge lines.
Figure 5.2: Comparison of traditional computer graphics techniques and techniques for creating technical illustrations.


Phong-shaded image.	Image with edge lines only.
Figure 6.1: Phong shading versus edge lines.


Image with new shading without edge lines	New Shading with Edge Lines.
Figure 6.2: Edge lines


Figure 6.3: These computer generated images were produced by using the illustration convention of alternating dark and light bands, to convey the metallic material property of the two images. The convention is rather effective, even for an object that may not naturally be metal.