rll WoRBWARE*GAME AND GRAPHICS POBICcx iS MObEUNG BASIC AND ADVANCED TECHNIQUES MARIO RUSSO 7)Polygonal Modeling: Basic and Advanced Techniques Mario Russo Wordware Publishing, Inc.Library of Congress Cataloging-in-Publication Data Russo, Mario. Polygonal Modeling : basic and advanced techniques / by Mario Russo. Pp. cm, Includes bibliographical references and index. ISBN-13: 978-1-59822-007-0 ISBN-10: 1-59822-007-1 (pbk.) 1. Computer graphics. 2. Geometrical models~Data processing, 3. Polygons. |. Title. T385,R865 2005 006.6'93-de22 2005025871 cP, © 2006, Wordware Publishing, Inc. All Rights Reserved 1100 Summit Ave., Suite 102 Plano, Texas 75074 No part of this book may be reproduced in any form or by ‘any means without permission in writing from Wordware Publishing, Inc. Printed in the United States of America ISBN-13: 978-1-59822-007-0 ISBN-10: 1-59822-007-1 109876543 0510 3s Max isa registered trademark of Autodesk, Inc, in the USA and/or other countries. Maya isa registered trademark of Alias Systems Corp. in the United States and/or other counties, ‘Other brand names and product names mentioned in this book are trademarks or service marks of their respective companies. Any omission or misuse (of any kind) of service marks or trademarks should not be regarded as intent to infringe on the property of others. The publisher recognizes and respects all marks used by ‘companies, manufacturers, and developers as a means to distinguish their products. This book is sold as is, without warranty of any kind, either express or implied, respecting the contents of this book and any disks or programs that may accompany i, including but not limited to implied warranties for the book's quality, performance, merchantability, or fitness for any particular purpose. Neither Wordware Publishing, Ine. nor its dealers or distributors shall be liable to the purchaser or any other person or entity with respect to any liability loss, or damage caused or alleged to have been caused directly or indirectly by this book. All inquiries for volume purchases of this book should be addressed to Wordware Publishing, Inc., at the above address. Telephone inquiries may be made by calling: (972) 423-0090Chapter 1 Polygonal System Basics Customization Mastering the tools of your 3D application is only a small part of the battle. It is highly recommended that you customize the keys of your application since the default keys often aren’t the perfect com- bination for you. When you customize the keys and shortcuts, you fit the keys to match your needs and the program adapts to you as much as possible in the most ergonomic and efficient way. 3D appli- cations tend to have enormous numbers of commands, tools, editors, and options, and if you don’t map shortcuts for your key- board, it can stress your muscles and slow you down. Think about the modeling process: You pan the camera, rotate the camera, zoom the camera, scale, move, rotate, zoom again, pan — all done using the mouse. Customization also allows you to change colors, the dis- play, and many other preferences. Of course, not everything can be changed or mapped to your keyboard or customized in some way, but the most important things (depending on the software you are using) probably are. Here are a few suggestions for customizatio > Map the viewport controls for your keyboard like this: A= Pan S = Rotate D=Zoom F = Zoom to selected object > Map the manipulation controls like this: Q = Select W = Move E = Rotate R = ScaleOrme Cur Aad Layouts The choice of a good layout will vary depending on your project. If you are an animator, it is desirable to have an animation editor (such as a graph editor, dope sheet, etc.) in a viewport to quickly access when needed. There are a number of ways to set up your layout, as shown in Figure 1-1, but for modeling purposes, you will have four standard viewports: Top, Front, Side (right or left), and Perspective. Figure 1-1 Figure 1-2: The standard layout for modeling: Top, Front, Side, and ive views.Ge Teoh an Med gfe ( LS) Ud 3D Polygonal Surface Before discussing modeling features, attributes, and techniques, you must understand what a 3D surface is. The mesh you will be working on is a group of vertices and edges that form polygons. The 3D surface is generated by advanced mathematic algorithms that put a “skin” over the mesh so you perceive a three-dimensional shape. Note that what you see in the viewport of your software is a “rough draft” compared to a rendered image. When you move your mesh inside the viewport, it is redrawn in real time. Figure 1-3: A wireframe render. Figure 1-4: Left, real-time surfacing. Right, final rendering of the same sphere with the same number of polygons.Ome CUS acu rad Space Alll of the 3D elements — polygons (meshes), NURBS surfaces, particles, lights, cameras, and helpers/constraints — exist inside a 3D world based on Cartesian space. When working with 3D model- ing in any software, you will be dealing with a three-dimensional world. This seems pretty obvious in theory, but in practice it means a lot of training, thinking in 3D, and in-depth knowledge of using the tools and accessing the viewports of the software that you are using to model your mesh. Every manipulation inside a 3D world must be planned inside a Cartesian system with three axes: X, Y, and Z. Figure 1-5: 3D space. In most cases, 2D software uses X and Y coordinates. Figures 1-6 and 1-7 show the difference between an orthogonal view (2D) anda perspective view of the same elements in a space. Figure 1-6: Orthogonal view (2D).Axis OTT me Cur SAUNT Tid NU Figure 1-7: Perspective view (3D) In 3D, all three axes are used to model, light a scene, animate, and more. The axis is probably the most used feature when you model and/or manipulate an object. If you want to move a vertex, edge, polygon, or entire object, or maybe scale it down or up, or rotate it, you will be accessing one or more axes. An axis is usually represented by arrows in the color red for X, green for Y, and blue for Z. Some pack- ages like 3ds Max and Maya have a different gizmo for each operation (move, rotate, or scale) so that when you turn on the tool you know how the operation will perform. z Figure 1-8Cae aac Polygonal Structure ‘The polygonal structure consists of three basic sub-elements: ver- texes, edges, and faces or polygons, as shown in Figure 1-9. 4 2 Figure 1-9: 1 indicates a vertex, 2 indicates an edge, and 3 indicates a polygon composed of two faces. As you can see, the cube in the above figure has eight vertices, 12 edges, and six quad polygons. The edges connect the vertices and form an n-sided polygon. In this case, we have a four-sided polygon, or quad polygon made with two tris. In 3ds Max you can select the Border sub-object fo select all the ‘edges around an open hole of your mesh. Polygonal Attributes Polygonal attributes represent the properties of the mesh, how it will look, how you will deal with its elements, and more. Among these attributes are Name and Display. Name When modeling complex objects or a scene, it is very useful to name your objects when you create them. If you model a bolt and you are working on an object that uses a lot of bolts, you may want to name them bolt000, bolt001, etc., so you can select all of them. from a list and apply one texture to all of them with one command. Naming your objects is also useful when you have to find an object in a complex scene and for general organization.eee a Saeed Name your objects as soon as you create them, and be sure to give them a descriptive name. Finding and renaming objects after the fact can be a cumbersome process. Display Display defines the mesh aspect in your viewport. It is common for most 3D applications to have the display options shown in Figure 1-10. Coe Figure 1-10: From left, wireframe, edged+facets, edged+smooth, and smooth without wires display options. Wireframe display disables the 3D surfacing and shows only the wires, or the mesh. Options related to backface wires or backface cullings will display properties of the wires that would not be seen if the viewport display were set to shaded view. The back sides of these faces will receive the same shading as the front sides. Backfacing culling usually means that the back sides of the faces won't be visible. Figure 1-11 shows a sphere with and without backface culling. Figure 1-11: Left, wireframe with backface culling off. Right, wireframe with backface culling on.eee Cuan rad Selecting When modeling, you can often select the whole mesh or sub-ele- ments (vertex, edge, face/polygon) of the mesh. Depending on the package you are using, the selection can be made in several ways. Whatever color scheme/user preference you choose for your viewport, keep in mind that the colors for the selected sub-element and the non-selected sub-element must be easily discernible. Figure 1-12: From left, vertex selected, edge selected, and polygon selected. The lessons and examples in this book use the selecting scheme shown in Figure 1-12. Selected vertices are shown as black dots, selected edges are black lines, and selected faces are white with black borders. Colors and Shading For modeling purposes it is recommended that you use a neutral color, like gray. However, if you wish, you can paint your mesh any color you like using the material/texture/shade editor. Depending on the model on which you are working, you may want to change the shading for a more appropriate look. It’s up to you to choose what best fits your needs. INDIE Git) San ame | If you are using Maya, the UV will appear as part of the sub-element of the mesh, along with vertex, edge, polygon, etc. UVs are used for texture mapping, not polygonal modeling, so they are not discussed in this book.eae ura C rid et Figure 1-13 Appropriate shading sometimes makes it easier to detect mesh errors like those shown in Figure 1-13. There are a few unneces- sary vertices near the left eye that cause a modeling imperfection. Adding a specular value to the shader makes it easier to detect the problem and fix it. Normals According to the 3ds Max user reference, a normal is a unit vector that defines which way a face or vertex is pointing. The direction in which the normal points represents the front of the outer surface of the face or vertex, which is the side of the surface that is normally displayed and rendered. This means that when you are not using a “force two-sided” option and the normal is flipped to the other side, you will not see the surface. Game programmers use normal manip- ulations for hardware acceleration and pixel shader effects, but for modeling purposes what really matters is that the normals show us the visible surface. It’s where we'll be working. An extruded face or edge follows the polygon’s normal.See CC Figure 1-15: Notice the flipped normals on the right. External polygons are invisible and internal polygons become visible. Polygonal Creation There are a number of ways to model an object. Most polygonal meshes start from a primitive, a polygon operation with splines, or a single polygon. Starting from a primitive is the most common way to create polygons. Complex shapes can be made starting with a box (box modeling), but there are other primitives that you can start from, like a sphere, a cylinder, a torus, or a plane. Figure 1-16: Primitives. You can also start with a spline (or curve) and then apply some mod- ification tool, like rotation, extrusion, or loft, to generate a new mesh.lea An ed Figure 1-17: Starting from a spline. By starting from a single polygon you can obtain full control over your mesh if you correctly plan the topology of the object. When you start with a single polygon, it is common to progress the model- ing by extruding the edge of the polygon until you get a shell of faces, as shown in Figure 1-18. Figure 1-18: Starting from a single polygon and then extruding the lateral edges three times. Figure 1-19: Selecting the bottom edges and extruding them a couple of times gives us @ quad polygon surface that started from a single polygon. With this method it is possible to create complex inorganic forms, as well as organic ones. With a good blueprint, you can use this method to model cars or any other vehicle.See Cece Operational Tools Operational tools are commands or modifiers that perform an opera- tion using two meshes, two lines or curves, or a single line and generate a new mesh as a result. Using operational tools can save time in many cases. For example, if you need to model a vase, using a Lathe operational tool is much easier than box modeling. If you need to model a cable, you can certainly make it by extruding a cyl- inder, but lofting is much quicker and more accurate. Let’s take a look at some of the most common operational tools available. Extrusion Extrusion is a modifier that picks a shape and raises a mesh from it that adds depth. See Figure 1-20. The process described as extrusion in this book is referenced as loft in Maya and is mostly for NURBS surfaces that are later converted to polygons. Figure 1-20: Using the Extrusion tool. Lathe The Lathe tool creates a 3D object by rotating a curve or line. When using the Lathe tool, the pivot point of the object must be centered correctly (the + in the following figure). If the pivot is not centered correctly, it may cause undesirable results. The Lathe tool usually allows you to shape the rotation in all three axes (X, Y, andSeat id Figure 1-22: Using the Lathe tool in the X, Y, and Z axes. Loft Loft creates a 3D object that extrudes one or more shapes along one path. Figures 1-23 and 1-24 show examples of using the Loft tool. The process described as loft in this book is referenced as extru- sion in Maya and is mostly for NURBS surfaces that are later converted to polygons. Figure 1-23: 3D object generated by lofting with two splines: a shape and a path. The figure on the right shows a slight decrease of the loft scale after the operation. 13Ome CaS acu Figure 1-24: 3D object generated by lofting with three splines: two shapes (circle and six-pointed star) and the path. The circle is lofted to one half and the star is lofted to the other half. Distribution Over a Surface Automatic placement of an object over other surfaces is the fastest option when you need to model some peculiar objects or scene, such as setting stones in a modeled or fractal-generated terrain in a random way, but you don’t want to or can't place them one by one. NOTRE Rak ET a A be ‘Maya has an advanced particle system that allows you to per- form a huge number of scatter effects over an object surface and animate it. Figure 1-25: Automatically placing objects.eae Ca anata Boolean Operations Boolean polygonal operations aren’t popular among 3D modelers because they reconfigure your mesh and often this configuration isn’t topologically correct. This means that when you smooth your Boolean-operated mesh, it will probably have some mesh flaws. Sometimes you don’t even have to smooth your mesh to see its flaws. However, in some circumstances, Boolean operations can be useful (otherwise, they probably wouldn't be included in most 3D packages as a tool). The golden rule is: If you can avoid using Boolean operations, avoid them. But if you can’t, be sure to do everything you can to check and correct all the flaws in your mesh. See Chapter 3 for more information about Booleans. A Boolean operation is one that, given an A object anda B object (as shown in Figure 1-26), creates a new shape. Boolean operations include subtraction, union, and intersection. B Figure 1-26 B-A A-B Figure 1-27: Boolean subtraction. _Figure 1-28: Booleon subtraction.eeu ec @ BD Figure 1-29: Boolean union. Figure 1-30: Boolean intersection. Pass-Through Mode Another useful property you may set for your object is the “pass-through” or “x-ray” mode. Not all 3D applications have it, but in some circumstances it can be very useful, such as when you're modeling a complex robot with a lot of mechanical parts inside it. It is also useful when rigging your character or in a situation where you need to see through a shaded mesh. Figure 1-31: Pass-through mode activated for the sphere. You can see the cube and the cone through the sphere.fe ae eau Global Deformations In this book, the polygonal tools that affect the mesh globally by a cage, lattice, or similar resource are classified as “global deformations.” Lattice Lattice is a cage that affects the whole shape by manipulating the cage’s vertices. In Figure 1-32, the image on the left shows a base cylinder without any modification, the center image shows the applied lattice, and the right image has some lattice vertex tweaks and scaled down top vertices. Figure 1-32 Figure 1-33 shows the same cylinder as Figure 1-32 with extra extrusion to get a more interesting shape. The image on the right shows the cylinder with the same lattice operation plus two levels of subdivision. Ud Figure 1-33Se ea c Twist ‘Twist operations are self explanatory. They twist the object, gener- ating very interesting results, but most cases require a good polygonal resolution to give smooth results. The following image shows the modified cylinder from the above exercise with different ‘Twist parameters. Figure 1-34 Figure 1-35 shows a basic line extrude with a high number of segments. Figure 1-35 Figure 1-36 shows the basic line extrude from Figure 1-35 twisted to model a rope.ee ee ea Seen Bend is another self-explanatory tool. It deforms the mesh accord- ing to the angle in the parameter. Figures 1-37 to 1-39 show several examples of using the Bend tool. Figure 1-36 Bend Figure 1-37: Bending in the Z axis. N ni Figure 1-38: Bending in the Y axis and in the X axis.Cee eu acne ed NOOO Figure 1-39: Earlier modeled rope, with 360 degrees of Bend. Polygon Reduction 20 Every object in your viewport is computed to display in real time, which requires a fast CPU, but even with a fast computer the high number of polygons starts to slow everything down. Sometimes a high-resolution poly for distant objects is not needed; in this case, polygonal reduction tools are the best choice to increase the speed and make the objects in your scene lighter. Figures 1-40 to 1-43 show the results of using polygonal reduction tools Figure 1-40: Left, original object with 109,170 vertices. Right, after the reduction operation there are 55,385 vertices.= f feat aren earg Basics yf | i a Figure 1-41: Left, 24,232 vertices. Right, 11,981 vertices. Figure 1-42: Left, 6,837 vertices. Right, 4,333 vertices. Figure 1-43: Left, the figure has 2,985 vertices shown in wireframe; center, the figure the same model without the wireframe; and right, the figure is the original mesh with 109,170 vertices. 21eee aca Organizational Tools In this book we use the term “organizational tools” for tools or fea- tures that allow the user to improve the organization of a complex scene or model in order to avoid losing time looking for specific objects and to optimize speed. Hide and Freeze Hide is used to hide objects in your scene. The objects are still there, but you can’t see or select them in your viewport; you will need to unhide them from the appropriate editor. The name for the Freeze feature may vary from application to application. Its main function is to lock the selected object for any viewport selection or transformation, such as translation, rotation, or scaling. Layers Layers have became an industry standard for computer graphics and multimedia. However, layers behave a little differently in 3D. In 2D applications like Photoshop or Flash, layers are used to deter- mine the depth of the elements inside the layers. In other words, layers determine what is under what. In 3D applications, layers are used like groups; if you add all the spheres to layerl, all the cubes to layer2, all the tori to layer3, then hide layer2, all the cubes (and any other objects) assigned to layer2 will be hidden (or reflect whatever properties that layer may have). It’s a quick way to set properties globally for all objects of a single group without having to select them individually. For character setup, for example, a technical director usually puts the bones of the character and the skin mesh in separate layers to make it easier to lock one and unhide the other. Conclusion 22 In this chapter we got a quick overview of the basics of the polygo- nal world, We covered the importance of customization, aspects of 3D surfacing, Cartesian space, and polygonal structure aspects and properties, along with some of the most common general deforma- tion tools and organizational features inside 3D packages.Chapter 2 Polygon Operations Before you start adding polygons to and cutting polygons from your shape, it is crucial that you understand how to manipulate its sub- elements because every modification and tool you apply to your mesh will affect the position and/or the number of vertices, polys, or edges of your mesh. The Move, Rotate, and Scale tools are self-explanatory; they move, rotate, and scale the selected sub-element. Each tool usually has its own manipulator gizmo. Move is represented by three pointed arrows, Rotate by a spherical gizmo, and Scale by a triangu- lar gizmo, as shown in Figure 2-1. Figure 2-1: From left, the Move, Rotate, and Scale tool gizmos. U & F Figure 2-2: From left, using the Rotate tool with selected vertices, the Scale tool with selected vertices, and the Move tool, moving the selected vertices backward after a scaling operation. 23eee Nake Cc iciry 24 These tools can affect sub-elements of an object, as shown in Figure 2-2, or the entire object, as shown in Figure 2-3. 4 aa s ~ Figure 2-3: From left, scaling in the X axis, scaling in the Y axis, and scaling in the Z axis. Another interesting point to mention when working with sub-ele- ments, or even an entire object, is the orientation of the gizmo in relation to the object. Most 3D packages will likely have a few options and allow you to choose the one most appropriate to your modeling. By World and By Local are pretty basic and the most common. When the By World option is selected, it doesn’t matter how much you have moved or rotated your object, the gizmo will be related to your viewport, as shown in the following figure. Figure 2-4: Using the By World option. When the By Local option is selected, rotating your object will cause the axis gizmo to also rotate and remain local to the object's internal axis. This can be very useful in some cases.eee ee eeu eae x Figure 2-5: Using the By Local option. Cutting Polygons The Cut tool is one of the most used tools for polygonal modeling. When you cut or manually divide the polygon, you add an edge. Every edge adds tension to the mesh, as we'll see in the next chap- ter. The procedure for cutting a polygon is specific to each modeling application. For example, since version 5 of 3ds Max there has been an indicator cursor that shows whether you are over an edge or ver- tex. In Maya, when you click on the edge and hold, you snap the cut starting point to the clicked edge; when you move the mouse you can reposition the starting point. 3ds Max 5 allows you to cut as many polygons as you wish with one cut. Maya 6.5 lets you cut one polygon at time. Figures 2-6 and 2-7 show examples of using the Cut tool. Figure 2-6: Click on the origin point (where you want the cut to start), then click on the destination point to generate a new line. 25eee ee euke ioe vue Figure 2-7: Here we have made a “+” cut in the face of a box and pulled the vertex forward. Extruding 26 The Extrude tool described here refers to extrusion of the poly- gon’s sub-elements, like a face, edge, or vertex. Face extrusion is oriented by its element normal, generating new faces along the side of the origin extrusion and extruded face. Edge and vertex extru- sion are similar to face extrusion. These processes extrude the selected sub-object, obeying the topology and using the object’s normal as the direction reference. Z Figure 2-8: Edge extrusion.See ee ake ec Figure 2-10: Face extrusion. Figure 2-11 shows extrusion operations combined with rotation operations. ou @@ Figure 2-11: Face selected, extruded, rotated a bit, extruded again, rotated, and extruded. Note that you can extrude as many times as you wish, and every time you extrude the face, new edges stay where you stopped the last extrusion. Most 3D packages give you several options for extrusion, including group extrusion, extruding by polygon, and extruding by local normal. With the Extrude by Group option, you extrude the selected polygons as a group. 27foe eee ke ce With the Extrude by Polygon option, the polygons will be extruded independently from each other. Figure 2-12: Extruding group. Figure 2-13: Extruding by polygon. With the Extrude by Local Normal option, the polygons follow the local normal of each polygon, but they stay together as a group. Figure 2-14: Extruding by local normal. 28ees ae ake aia Chamfer/Bevel Chamfer and Bevel can produce very similar results. In some appli- cations, however, the name related to its function may vary, causing confusion. Let’s take a look at the tools. It is very common to see beveling functions applied over 3D. text, giving the result shown in Figure 2-15. Figure 2-15: Beveling text. In 3ds Max, beveling applied to a box face will give us the result shown in Figure 2-16. Figure 2-16: Beveled box face in 3ds Max. Additionally, depending on the parameters that are set for this tool, the shape can vary considerably, as shown in Figure 2-17. 29eee Nake ct wiv Figure 2-17: Using the Bevel tool. 30 Figure 2-18: Selecting the top edges and applying the Chamfer command. Figure 2-19: Using the Bevel tool in Maya. The same result could be achieved by extruding the face up and then scaling the extruded face down obit. Note that the variety of options allows you to reach the same result in different ways; just choose the quickest or easiest method for you.Ce aera key ad Attach and Detach Detach is the operation to use when you have a mesh and want to separate parts of it into two distinct meshes. Figure 2-20: Detaching. Remember, every time you detach a piece of the mesh, each piece will be an independent mesh that can have different properties, tex- tures, and so on. Attach is a very useful operation that allows you to join two dif- ferent meshes and merge them as one. If you are modeling a full human body, you may want to model the head, the limbs, and the torso separately and then merge them into one single mesh. After the attach operation, all the parts of the human body you've mod- eled will be treated as one for any further operations you may apply. However, they do not necessarily have to share the same texture. You can merge two meshes and preserve the texture coordinates, shaders, and textures applied to each one. —P aE Figure 2-21: An attached mesh without merged vertices. After smoothing is applied, the meshes continue, disconnected from each other. 31eee Nake ce When you attach two distinct meshes, they will be a single mesh but will not have continuity. The attach operation does not merge the vertices automatically; you'll need to merge the vertices one by one. Ll al Figure 2-22: The attached mesh with the ve appl jously when the vertices at |. The smoothing is welded. Welding is probably the most common tool for fixing a mesh that has errors or matching two meshes after an attach operation. Welding works differently in different applications. In 3ds Max, you cannot weld a vertex that isn’t near a hole or a vertex that does not have an edge connecting through the targeted vertex. You can do this in Maya, but it may have undesirable results on your mesh. This book recommends “safe” techniques that you can apply using just about any 3D modeling program you wish. Figure 2-23: In Maya you can easily weld the vertices of the attached cylinders, but in 3ds Max you must delete the cap hole. 32Coe ae ued ea To use the Weld command, you select the matching vertices one by one and apply the command. See Chapter 3 for information on weld- ing meshes that do not have the same number of vertices. Figure 2-24: Using the Weld command. Figure 2-25: Welding two cylinders with the same number of sides. 33eae uke ea Duplicate and Mirror One of the most common tools in computer graphic applications is the Duplicate tool. This tool makes a copy of the selected object (usually with all the properties of the original object). Figure 2-26: Duplicating along the X axis. The Mirror tool duplicates the object and mirrors it across an axis. Figure 2-27: Half of a face mirrored in the X, Y, and Z axes. 34Perea eine Pivot Point ‘The pivot point is an attribute that represents the object's location and governs transformations such as rotation and scale. When the pivot point is centered, the scaling and rotation operations use the center point as their origin. When the pivot point is aligned to a cor- ner, the scaling and rotation operations use that corner as their origin. Figure 2-28: Centered pivot point. Figure 2-29: Pivot point with the origin in a corner. 35Chapter 2 - Polygon Operations Ahole is a missing face in your mesh. It’s not desirable in most cases, but if you know how to work with it, it may be useful. Figure 2-30: A hole in a mesh. Smoothing a Mesh with a Hole ‘A mesh that contains a hole can be smoothed. Figures 2-31 to 2-38, show different smoothing operations. Figure 2-31; The sublvsion computes the cop os a tension generator on this four-sidi ry edge contributes to the new mesh generated by the subdivision process aonren te right. 36Sree eae eioaS Figure 2-32: Left, a four-sided cylinder without a cap. Right, the same ylinder subdivided. As you can see, Catmull-Clark subdivision does a great job on quad polygon shapes, and the four-sided cylinder becomes a smooth cylinder ofter a few subdivision steps. Figure 2-33: Left, the cylinder with a cap, after subdivision. Right, the cylinder with a hole. Notice that the hole does not generate tension on the mesh when subdividing. Figure 2-34: To show how subdivision performs on a quad hole, we pick @ cube and delete the middle face on one side. 37OI) are 7+ [Oca Figure 2-35: Left, the same hole without subdivision. Right, with one level of subdivision. Figure 2-36: Left, two levels of subdivision. Right, three levels of subdivision. Figure 2-37: The same cube but with the hole distorted a bit. 38ae ee ue) eeu Figure 2-38: More levels of subdivision. Soft Selection Soft Selection, SoftModificator, DragNet — the name can vary from application to application but the purpose is similar: to select a sub-element and apply an influence for its neighboring elements. For example, if you move a vertex, the neighboring vertex will also be moved, depending on the Attenuation parameter of this tool. It’s like a magnetic tool attracting the nearest sub-elements of the selected sub-element. Figure 2-39: Using Soft Selection. Soft Selection tools are very useful for reshaping your objects and adjusting their proportions and volume. 39le ee uke) cay Mesh Errors 40 Mesh errors are any undesired artifacts or holes, or continuity errors on the mesh, It’s not uncommon to find errors on the mesh. a Figure 2-40: Left, an error mesh. Depending on the viewport shading, it may be hard to find out where the error is before subdivision. Middle, shading with o high specular level (subdivision level 1). The error is still not easy to find. Right, the figure has two levels of subdivision and it is more obvious now where the gap i S54 Figure 2-41: The mesh error in this figure is caused by two vertices that are 00 close. The solution is to merge them together. Figure 2-42: Note the problem on the chin.Leiitel aes 7* RO) cad ‘To find and correct mesh errors, smooth the object and carefully watch the geometry while searching for them. Mesh errors can pro- duce undesirable results at render time. 4Chapter 3 Polygon Subdivision Polygon Subdivision Theory Subdivision (also called smoothing) is a process for generating smooth surfaces by refining a polygonal mesh. The basis for subdi- vision can be traced back to the 1940s and ’50s when G. de Rham used “corner cutting” to describe smooth curves. Since then, many researchers and scientists have worked on this topic, but it was only in the mid-1970s that the practical use of subdivision theory as applied to computer graphics started to take a definitive shape. In this chapter we discuss subdivision algorithm theory and provide a number of examples. All of the subdivision methods described in this chapter are classified by three criteria: > Type of mesh: quadrilateral (quad) or triangular (tri) > Type of refinement: primal (vertex insertion) or dual (cut-corner) > Type of scheme: approximating or interpolating The mathematical procedures that define the subdivision process are called “masks.” In this chapter we show the visual representa- tions of the subdivision schemes for control points with the coefficients of the subdivision mask. Arbitrary topology means that the edges, vertices, and the whole graph formed by the mesh cage can be arbitrary; that is, the vertices of the mesh may be of arbitrary degrees. In arbitrary topol- ogy, we classify the valence of the vertices when 6 degrees as tri and when 4 degrees as quad. Ordinary vertices have a valence of 4, and extraordinary vertices have a valence other than 4. 43ee ae NC sca Figure 3-1: Two extraordinary vertices with valences of 6 and 3. Chaikin’s Algorithms for Curves (1974) In 1974, George Chaikin introduced one of the first algorithms for refinement (cut-corners). In contrast to his predecessors, Chaikin developed an algorithm that worked directly on control polygons. His new algorithm generated a new control polygon by cutting the corner. Ps Ps Figure 3-2: The initial, unrefined polygonal line with five P points: Po, Py, Pa, P,, and Py. 44Chapter 3 ~ Polygon Subdivision Figure 3-3: Refinement step 1. Chaikin’s method generates Q and R points, while cutting the corners add refinement to the line. Figure 3-4: Refinement step 2. Figure 3-5: Refinement step 3. When increasing the refinements, the curve gets smoother. 45ee eeu ia 46 Doo-Sabin Surfaces (1978) Daniel Doo and Malcom Sabin adapted Chaikin’s polygon refine- ment technique for the bi-quadratic uniform B-spline and developed anew procedure for surface generation. Doo-Sabin is a dual approxi- mating scheme for meshes of arbitrary topology. All the vertices have a valence of 4 after the first subdivision refinement. Figure 3-6 Figure 3-7: Left, an unrefined mesh. Center, vertices placed by the Doo-Sabin method and then connected. Right, the subdivided shape is shown in black and the original, unrefined mesh is shown in gray. Catmull-Clark Subdivision (1978) The algorithm proposed by Edward Catmull and Jim Clark is the most widely known and used. When the mesh is subdivided, new vertices (face points) are placed at the center of each unrefined face and in the center of each unrefined edge, and then new edges are created to connect these new vertices. It’s an approximating primal scheme.eee ae unt eta ‘These are the steps of Catmull-Clark subdivision: 1. The face points are created by averaging positions of the poly- gon’s unrefined vertices. 2. Edge point locations are calculated by averaging the center point of an unrefined edge and the locations of the two new adjacent face points. 3. The positions of the old vertices are reconfigured using the following equation: -2,,2R,, Sin-3) een Figure 3-8: Variable @ represents the average of the new vertices (face points) surrounding the unrefined vertex. Variable R represents the average of the edge’s midpoints that share the unrefined mesh. Variable S is the original, unrefined mesh. Variable n is the number of edges that share the old unrefined vertex. a 2 i6 i6 1 1 a 4 3 bk B 8 1 L 1 1 sar a" iS % 4 4 Figure 3-9: Left, a mask scheme for an edge vertex. Right, a mosk for a face vertex. 1 1 oe $ = 2 > 8 4 8 Figure 3-10: Left: « mosk for boundary odd vertices. Right, a mask for even vertices. 47Cee a enn a 1 16 16 5 2 B 8 2 2 16 16 Figure 3-11: Leff, « mask for extraordinary vertices. Right, « mask for interior even vertices. Catmull-Clark is an excellent scheme for ordinary vertices (valence of 4, quads) but also works with valences that are greater than or less than 4. The following figures demonstrate the Catmull-Clark subdivi- sion scheme. A Figure 3-12: An unrefined cube primitive seen from the top view. The face points are generated and placed using the Catmull-Clark subdivision method. 48ers eeu hia LAS SA Figure 3-13: New vertex points (F) generated from the (E) points (see Figure 3-11). These new vertices are then connected, forming the subdivided shape. Figure 3-14: Subdi Figure 3-15: Refinement of subdivision example. 49fe ee ST hor Loop Subdivision (1987) The Loop subdivision method is an approximate primal scheme for triangular meshes proposed by Charles Loop in 1987. The Loop method can be applied to triangular arbitrary polygonal meshes. ‘The following figures show the masks for the Loop subdivision scheme. coke eof ce Figure 3-16: Interior rules. Left, rules for odd vertices. Right, rules for even vertices. 1 es g a =a = 8 4 2. 2 Figure 3-17: Special rules for creases and boundaries. Left, rules for odd vertices. Right, rules for even vertices. Figures 3-18 to 3-21 show Loop subdivision in action. Figure 3-18: Unrefined mesh cube with 12 (tri) polygons. 50fea he eT cy Figure 3-21: Third level of subdivision, 768 faces. 51ore ee aetna Butterfly Subdivision (1990) Butterfly subdivision is a method proposed by Dyn, Gregory, and Levin and is primal and interpolating for triangular meshes. In the original subdivision scheme, artifacts often appeared on the mesh after the subdivision interpolation. In 1996, Denis Zorin’s Modified Butterfly Subdivision scheme improved on the Butterfly method, solving the problems of the old scheme yet retaining the simplicity of the original scheme and improving the smoothing. Figures 3-22 to 3-25 show the subdivision steps for the Butter- fly scheme using a triangulated cube. Figure 3-23: First level of subdivision, 48 faces. 52eee unt cea Figure 3-25: Third level of subdivision, 768 faces. Kobbelt Subdivision (1996) The Kobbelt subdivision method is interpolating and primal, and works on quadrilateral meshes. The difference is that vertices are fixed during the interpolation process and the rules for edges and faces are referenced to the unrefined mesh. 53ee ae hori Figure 3-26: Kobbelt's scheme for computing a vertex adjacent to an extraordinary vertex. ae 3, es a Pace ene earn Rey ae ah 16 | (6 1% 16 ge Figure 3-27: Left, a regular mask for edge, crease, and boundary vertices. Right, a regular mask for a face vertex. Continuity Continuity can be defined as the smoothness of an object’s edge. Computer-generated models should not contain any faceted poly- gons, gaps, or hard edges in the final result (unless they are deliberate). Some modeling formats, like Maya's subdivision sur- faces or NURBS, have implicit continuity, but balancing the weight of the mesh must be carefully planned when using polygons. 54Figure 3-28: Left, bad continuity. Note the facets at the top of the head. Right, after one step of subdivision. There are still some artifacts. Figure 3-29: The same object with two levels of subdivision. We can almost see the facets in wireframe, but at render time they are nearly imperceptible. Subdivision and Wire Tension Most 3D applications use Catmull-Clark subdivision. It is the indus- try standard for dealing with polygonal geometry. As this subdivision algorithm works best with meshes with ordinary verti- ces (valence 4, quads), effort has been made to keep the methods and techniques quad as much as possible. When that’s not possible, non-quad polygons are used in areas of less deformation. The difference between a cube polygon with six faces (quad) (see Figure 3-30) and a cube with 12 faces (tri) (see Figure 3-31) is very clear after some Catmull-Clark subdivision: The edges across the tri faces generate unnecessary tension, giving a deformed look to the cube during the subdivision process. 55eee CaS bra woee Figure 3-30 Figure 3-31 © & Figure 3-32: Left, final result of cube subdivision of six quad faces. Right, final result of cube subdivision of 12 tri faces. Figure 3-33: Subdivision process of a pentagon cylinder mesh. 56er eee Ne Starting from a Primitive “Imagination is greater than knowledge,” Einstein once said. Using this as a golden rule when starting polygonal modeling from a primi- tive will bring you good results. Knowing your primitives well and planning what they can become is a very good starting point. Don't think about primitives in terms of what they are; think about the possibilities. A box can be a pipe, a cylinder can be a vase, and a cone can be a flower. When starting from a primitive, the sky is the limit. Figure 3-34: We extrude the cube a few times. Each line adds more tension. Figure 3-35: The cube becomes a cylinder. We extrude the top polygon, scale it down, and extrude in. 57Oe Nuc 58 Figure 3-36: After some subdivision we get a tube made from a cube. Figure 3-37: With a rigid cone pri you can scole up the body vertices ‘and push down the top vertex a little bit, making the cone smoother. Then scale down the whole object. RS py Eo Figure 3-38: A cone becomes a flower.Ce ae a a) Giving It the Shape You Want Creating the desired shape when modeling polygonal objects is all about understanding how subdivision works. The desired shape often depends on careful cuts and dense and structured wires. A well-planned mesh can give desired results without excessive hours of cutting and reshaping. Here we discuss some of the most impor- tant aspects of planning complex geometry: edge flow and edge loops. Edge Flow The edge flow represents the actual structure of your model and how the topology must be constructed in order to work correctly according to the anatomy of the model or object when animated and to avoid unwanted gaps, hard edges, or subdivision artifacts. Notice in Figure 3-39 that the lines of the male model (at left) follow the flow of the human anatomy to provide the correct tension and deformation of the muscles after the subdivision. The female model's abdomen (at right) is not correct from the muscle structure standpoint, but the base mesh as shown can provide a good struc- ture for an abdomen without bumpy musculature. Figure 3-39: Edge flow. 39eee ON na Edge Loop Edge loops occur when the end of the line encounters the begin- ning, forming a loop. Figure 3-40: Edge loops. The same thing occurs when a line contours a model and returns to its origin. Subdivision and Iterations oe 60 Polysmooth is a term for smoothing polygonal surfaces, and is also called subdivision, which was discussed earlier in this chapter. The iterations correspond to the number of subdivisions, meaning the higher the iteration value, the higher the number of polygons and the smoother the surface. One secret to working with a 3D scene is to manage the weight of the elements in the scene. This means, in geometry terms, that you need to carefully watch the number of visible objects in your viewport and the density of the mesh (num- ber of polygons). Sometimes a higher number of polygons (high iteration value) is necessary for a good general resolution of the object (better continuity). The following figures are two examples of smoothing with changing iteration values.ore eee chr Inorganic Modeling Inorganic modeling is not the easiest approach for polygonal model- ing if you need precision. However, polygons can do a great job if you plan the mesh and use it wisely. Here we discuss techniques for successful inorganic modeling. Soft Edges The look of edges is very important in the creation of realistic inor- ganic objects. Cut objects are among the few real-life objects that have sharp edges, similar to those we get when creating box primi- tives in 3D applications. In 3D, the corners of inorganic objects must have at least a hint of a soft edge to improve the look and real- ism of the objects (the shading also benefits from edges that are not too sharp). In this section we describe how to control the creation of hard and soft edges, using Figures 3-43 to 3-52 as examples. 61eae aS hd we oa 62 Figi 43: There is not enough tension to hold the box’s shape after subdivision. yo QY Figure 3-44: Adding edges improves the tension and holds the box shape. Figure 3-45: the tension is higher and the border gets sharper; when the extra edges are farther away, the border gets rounder.ere ee ae en Figure 3-46: In this example we see clearly how the sul smoother look to the box. Figure 3-47: The closer the edges, the sharper the border. When there are extraordinary vertices (valence other than 4, which is the case of the vertices on the top border of the cylinder in Figure 3-48), the subdivision algorithm has to perform operations to com- pensate for the lack of vertices. Figure 3-48: In this case, where there is a valence of 3, there is one vertex missing that keeps it from having an ordinary valence of 4. 63eee Cant ia ISS ‘| a Figure 3-49: When we see the result without the wireframe, some ripples oppear on the smoothed border. ‘To correctly generate the cap polygons in order to avoid unwanted tension during subdivision, select the top polygon, extrude up, scale down, and move down to align with the top edges, as shown in Fig- ure 3-50. Figure 3-50 Repeat the operation. : we! Figure 3-51 64CO ae ue he Then you can select all the edges and weld them together, as shown in Figure 3-52. Figure 3-52 Now you have a nice top that you've made manually. Bump Modeling circular edge loops by hand is the safest way to carve or bump polygonal objects and reduce mesh errors. When you cut an eight-sided edge, you have the opportunity to plan how the mesh will behave because you can cut, remove edges, and merge vertices as much as you need. First we cut an “X” in the top of the box, then cut a “+” to give us eight intersecting lines. Then we draw a poly-circle in the mid- dle, as shown in Figure 3-53. Figure 3-53, Select the polygons inside the circle and extrude it a little bit, then extrude again. This will generate tension in the base. Extrude again, a little higher. 65ee ee a Nee 66 Figure 3-54 After the extrusion, scale down the top polygons. You may want to cut near the top of the extruded object to improve the tension at the top. Figure 3-55 Notice the differences between the two images in Figure 3-56. The one on the left represents the extruded shape, and the one on the right is the same shape with one of the base edges removed. Figure 3-56ere eee a iia Circular Depression You can also create a circular depression. Start by creating a rectan- gle and slicing the polygon with the Cut tool, as shown in Figure 3-57. Scale up this rectangle to get a more elliptical hole when we extrude down. Figure 3-57 With the face scaled we can cut the ellipse inside the rectangle, cut- ting from vertex to vertex. Select the circular polygon and extrude down to generate new tension where the extrusion stopped. Figure 3-58 By making the hole manually, we can set the shape and the border in many ways. 67eee Cunt hcl 68 Figure 3-59 The above step gives us the smoothing shown at the left in Figure 3-60, but we can chamfer the lower edge and get a much sharper result. Figure 3-60 Figure 3-61 shows the results of creating a bump and a circular depression after three levels of subdivision and shading with a high specular value. — Figure 3-61ee eens ein Cracks Cracks are an important part of inorganic modeling. With polygons you can set a variety of cracks in just about any type of object. Fig- ures 3-62 to 3-67 show how to create a crack in an object. Starting with a 12-sided cylinder, we approximate three edges, as shown on the right in Figure 3-62. Notice that this modification changes nothing when rendered. Figure 3-62 Pushing down the middle vertices makes a very smooth crack appear, as the tension there is very low. Figure 3-63 69eee hic 70 Figure 3-64 shows two edges after a chamfer operation. The tension generates a sharp crack. Figure 3-64 All the edges are involved now and there is more tension. Figure 3-65 Figure 3-66eeu Chee ‘The vertices are moved closer and the crack gets thin. Figure 3-67 Body Cracks In this section we discuss how to efficiently make holes in your mesh and get a nice, smooth mesh without errors. Create a rectangle structure using the Cut tool and then select the middle edge and chamfer it. Instead of one edge, you'll have two. Cut in the middle of the center polygon as shown at the right in Figure 3-68. Figure 3-68 Now we'll create the rounded corner structure. Make cuts as shown at the left and center in Figure 3-69, and notice everything outside the crack is as much quad as possible. Select the interior polygons of the crack, extrude down a bit, and then extrude again and again. ‘The edges generated by the first extrusion create good tension on the inside part of the border’s crack. 7Se ee ea 72 Figure 3-69 ‘The general structure is done. However, you may also want to cut an extra border outside the crack border loop, as shown in the cen- ter in Figure 3-70, to improve the local tension. Figure 3-70 Thickness Primitives and polygons by default don’t have any thickness. If you delete half of a polygon sphere, you'll notice absolutely no thick- ness. 3ds Max 6 has a tool named Shell Modifier that allows you to. give thickness to your object. In this section we show how to add thickness to your mesh without using any special tool or script. First, pick a sphere and delete half. Duplicate the object and scale it down a bit.