This week I talked about graphics standards and the exam.

COMP4431 will also demo their games, so if you are thinking of doing COMP4431 next year, you should definitely attend.

More details here.

My slides (ODF) Slides in PDF

Some past exams.

Cheat sheet (this is included at the back of your exam paper).

Standards are normally promoted by organizations such as the International Standards Organization, however many of the standards used in computing are de facto standards, created and promoted by a particular company.

File Formats

There are very many graphics file formats out there. Wikipedia has a good summary of Encyclopedia of Image File Formats. Another useful site is Wotsit’s, which has links to all kinds of formats.

Image File Formats

An image is just an array of pixels so we can just store them in a file.

GIF

A GIF file uses 8 bits per pixel, compressed. It was the first file format supported by web browsers, so is very popular. Unfortunately it was patented.

PNG

Developed as replacement for GIF, PNG is patent-free and supports 8, 24 and 32 (RGB plus 8 bit transparency) bits per pixel. It uses a better compression scheme than GIF.

JPEG

GIF and PNG use lossless compression. JPEG compresses files much more by throwing some information away, that is the original image cannot be created exactly, but the difference is not noticable to humans. JPEG 2000 is designed to be the successor to JPEG.

Movies

MPEG

MPEG uses lossy compression similar to JPEG. It also compresses movies by only sending the differences between frames.

Animated GIF

This is just multiple GIFs stored in one file. It is popular because it was the first movie format supported by web browsers.

MNG

is to animated GIF what PNG is to GIF.

MPEG 4

This extends MPEG to multimedia content. It adds things such as 3D objects (like VRML) and capabilities such as letting the viewer change their point of view.

Object Formats

Instead of encoding the pixels coming out of the far end of the graphics pipeline you can encode the 3D (or 2D) models that go into the pipeline.

SVG

SVG (Scalable Vector Graphics) is an XML based description for 2D graphics.

PostScript

This is a page description language developed by Adobe originally intended for specifying the pixels to fill on laser printers. 2D Shapes are specified by Bezier curves. One interesting feature is the language is a simple programming language, thus allowing easy extensibility.

VRML

Intended to allow 3D environments to be put on the web, VRML describes 3d geometry, appearances and behaviours.

X3D

X3D is the next generation for VRML, combining it with the XML (the next generation replacement for HTML).

APIs

Application Programmer Interfaces are the means by which programmers access library functions.

OpenGL

OpenGL provides 2D and 3D graphics functions, including modelling, transformations, color, lighting, smooth shading, as well as advanced features like texture mapping, NURBS, fog, alpha blending and motion blur. It is available for Windows, MacOS and Unix systems and can be called from C, C++ or Java. Many 3D graphics cards provide OpenGL support.

Direct3d

This is Microsoft’s attempt to duplicate OpenGL in an API they control.

Java 3D

This is a higher-level API than Open-GL. Essentially you create and modify a scene graph, an object representing the entore scene to be rendered.

Open Inventor

This provides similar functionality to Java 3D, but is built around C++ instead of Java. VRML is based on Open Inventor’s file format.

OpenSG

OpenSG is a portable scene graph system to create realtime graphics programs.

OGRE

Open Graphics Rendering Engine: a open source 3D graphics engine that has been used in several games.

Question 1

This question was contributed by Daniel Lambert, age 8.

How can you tell that the image on the cover of the text book is not a photograph of a real scene?

If it was a photo, the camera would be visible in the reflection on the mirrored sphere like this.

Question 2

The teapot is actually modelled mostly using Bezier patches. The base and lid can be modelled as surfaces of revolution, and it’s also possible to model the spout using extrusion, but this is tricky. It’s actually done using Bezier patches. The surface is usually a metallic or at least a smooth surface. Hence we would probably model it as a highly specular (i.e. with a very high phong exponent (maybe, say, 40). The colour depends on the type of teapot, but if it’s the traditional gold, then that’s fine.

A barrel can be modelled easily as a surface of revolution. Another way it’s possible to imagine making a barrel shape is to use an ellipsoid from its implicit surface definition, then using constructive solid geometry, remove the top and bottom of the ellipsoid. A barrel has wood panelling. Wood is generally a matte texture, so not very much specularity — unless, of course, it’s polished or lacquered. Barrels are neither of these. To model the texture, we could use a normal texture map, but it might also be worth considering procedural texture mapping for more realistic effect. However, since it is likely to be used in a real-time game, probably we will go for a texture map. We could also use bump mapping to model the “grooves” between the wooden pieces which are assembled into barrels.

A shot barrel is the above surface of revolution, but you can use CSG to mark the hole in it as a cylinder. If the cylinder that shows where the weapon went through is C, and the barrel is B, then the CSG expression for the shape would be: B - C The surface is not much different from an unshot barrel. The only thing to be concerned about is visibility of internal textures and the texture at the point where it goes through the wood. It would look fake if we used the same texture for both the surface and the bit of the hole, but it probably doesn’t make too much difference.

Exhaust pipe can be accomplished in many ways, but the easiest is to use extrusion and have a cross section of the tube and have different scale factors at the beginning and end. Again, exhaust pipes are usually shiny chrome, so highly specular surfaces are good. If we really want to get that “chrome” look right, then we need to use an environment map so we get realistic reflections off the pipe. If we really wanted to model bits of dirt and stuff as well, we could apply a partially transparent texture.

Marble pillars? This is just straight combination. The cylinder in the middle is done either as a SOR or extrusion. The base is modelled using polygons. Marble is a situation where you really stand to gain from the use of solid textures to get that really nice marble effect. Plus, it would look very natural at the point where the base and the column meet: the textures would match so it would look pretty natural. Polished marble is pretty specular, so you may again need a Phong exponent, maybe not as high as for metallic surfaces, maybe 10.

Carapace is best modelled using Bezier patches, or their big brothers, NURBS patches. The carapace is probably partially specular but not too specular. It could have different colour patches on it, especially in the case of an alien.

Question 3

I am not providing a sample answer to this question. I do not want students repeating it verbatim as an answer to vaguely similar questions on the exam.

Question 1

This question was contributed by Daniel Lambert, age 8.

How can you tell that the image on the cover of the 2nd edition of the text book (click to enlarge) is not a photograph of a real scene?

Question 2

You are applying for a job as a computer graphics expert working for Microforte (an Australian games company). In their employment exam, it asks what modelling and rendering techniques you would use to model the following:

• A teapot (this is required by tradition).
• A barrel.
• The same barrel after it has been shot through with a weapon, leaving a circular hole.
• A fancy-looking exhaust on a flying motorcycle. The cross section stays the same shape, but it “flares” out at the end
• A marble pillar. The base and top are square, but the midsection is a circle.
• The carapace (i.e. the hard head covering) of an organic-looking robot like the hard shell of the evil alien in the Alien series of movies.

Discuss how you model the shape of these objects as well as how you would implement the surface of these objects. What techniques would you use to model the surfaces in the above cases?

Question 3

This question is from the COMP3421 exam in 2000.

SOCOG (Sydney Olympic Games Organizing Committee) wants you to build a virtual guide for the electronic information kiosks at the Olympic site. They want users to be able to look and walk around a 3D model of the Olympic site at Homebush Bay. They want people who have purchased tickets to be guided (in the model) to their seats so they can see how to get there.

Tell me (using diagrams where appropriate)

1. what hardware your system would use,
2. what software and rendering algorithms,
3. how you would model the Olympic site,
4. and how a user would interact with your system.