Previous Contents Next Mass Scenes Rendering Framework page 1 of 8

1. Introduction

One of the goals of current computer graphics is to produce images with as much realism as possible. To achieve this, there is not only the need to improve old (or invent new) rendering techniques, but also to enhance the scene. In the recent years hard work has been done to make the modeling tools as comfortable and easy to use as possible. An experienced graphic designer is capable of creating a 3D model of almost any shape he can imagine. However, what good is an excellent three-dimensional model if it is in the scene alone? To create a realistic-looking scene it is necessary to place many entities at proper places. It is the scene composition that often leaves in the spectator the biggest impression. A brief glance at what support is available to the scene designer to make a scene with many objects constitutes an opinion that a mass-scene creation and rendering tool would improve both the quality and the comfort of a 3D-scene design.

1.1 Typical mass scene

Typical scenes that have been considered during the design of the mass-scenes creation tool include: characters in an audience, a flock of birds, a lawn with blades of grass, trees in a forest, a hairy monster, stalagmites in a cave, cars on a parking place.

If the scene author has enough time and patience, he might be able to do a very nice mass-scene without any special instrument, but with a great effort. Current modelling tools provide only two techniques to clone an object: instantiating and referencing the particular object. Although they are quite good to model small quantities of objects, they are not sufficient for mass scenes. The aim of this paper is to make the scene creation work much easier - what can be made by the computer should be done by the computer without any painful, long lasting or boring human assistance.

1.2 Instantiating an object

Instantiating is one of the basic operations that are in almost all software products. It is often called copy & paste. This facility makes possible to create a lot of instances, to place them into the scene and later change their properties. Some modeling tools provide functions to make the placement automatically - in a row, on the circle, or even some more sophisticated layouts.

1.3 Referencing an object

Referencing is almost the same as instantiating. The only difference is that if an object is cloned as a reference, after making a change in the original object, all the references are changed too. This has a big advantage if there are changes to be made after the objects have been cloned. E.g. the author forgot to add eyes to a creature that has already been cloned hundred times. After adding the eyes to the master (original) object, all the slaves (references) receive their own eyes immediately.

These two techniques have one main disadvantage - all the changes to the clones are to be done by the designer. The computer only makes the clones that are exactly the same as the original. However, for a realistic look the instances must differ in some details.

Due to the lack of time or patience, the author of the scene alters only a scarce number of the instances. At the first glance the result seems to be quite nice and fulfilling the intention, but after a short observation the objects seem to be somehow periodical and disturbingly regular, the scene is not as realistic as should be.

1.4 Special modeling techniques

A lot of hard work has been devoted in the recent years to some special types of mass scenes and parametric descriptions of objects. The results have usually the same basic scheme - the object to be modeled is considered from two points of view, the first is how an instance of the object can look like, and the second is how the instance behaves in an area or as a part of a whole entity. Then a tool is developed that confirms the assumptions. An example of such paper is [1] where the authors studied plant ecosystems and then rendered scenes consisting of thousands of generated plants in amazing photo-realistic quality. Another example is the modeling of human hair [2] by making a physical model of human hair, and thus generating the whole head.

Also the movie producers have their own special modelling tools that are many times developed exclusively for a desired type of scene. For instance the famous company Pixar developed in 1998 for the movie A Bug's Life new methods for modelling and animating large crowds of figures, but they keep their technical information in secret.

Nevertheless, all these modeling techniques are quite specialized for a particular type of scene. This paper introduces a technique that is not dependent on the scene type; it is a tool that is applicable on most mass scenes. On the other side it cannot compete (in the image quality) with a very special modeling tool, but this small handicap is counterbalanced by the generality of the approach.

Mass Scenes Rendering Framework, Dušan Bezák, 1999-2001,