The RealSoft3D ray-tracing process - Part 1: Introduction

By David Coombes
d.g.coombes@btopenworld.com

RealSoft3D provides perhaps the most powerful raytracing engine available in any 3D rendering package. It's open-ended design allows the user access to the various stages of the ray-tracing process to control a render's appearance in an incredibly versatile manner. This allows incredible photo realism, cartoon-style shading, textured line-art, or any number of new styles without the need for expensive plugins or custom renderers.
To make the most of this incredible system you really need to understand how RealSoft3D generates images from the 3D data provided through the scene modelling and how it fits in with RealSoft3D's primary control mechanism, the Visual Shader Language (VSL)

This document is designed to introduce and explain RealSoft3D's processes. To help with explanation I'll use as example a simple scene consisting of a red sphere on a blue and white chequered ground, with a white wall behind. There is one light source from the front and left of the scene. The camera is placed to view the scene straight on with a perspective view and square aspect ratio.

The scene, showing the camera's view as dotted lines	The scene as rendered

This document does not assume prior knowledge of 3D computer graphics although it will assume knowledge of basic VSL as covered in the RealSoft3D manual. It will also make use of standard computer-graphic terminology such as pixels. Finally, some of the explanations are not accurate in relation to the mathematical mechanics of image generation or the exact order of events, but are deliberately used to aid understanding and help with visualizing the process. I hope in this document I manage to explain this such that you can really tap the power provided by RealSoft3D!

Casting Rays

First things first, RealSoft3D is a ray-tracer. This means that it creates images by tracing rays. A ray can be thought of as a straight line emanating from a source and proceeding until it hits an object. When tracing an image, a ray is cast for each pixel of the image. For example, let's render our scene as a 10 by 10 pixel image. For the first pixel in the top left corner, a ray is cast from the camera (see image on right ; the ray is the green line). The angle this ray is sent is determined by the camera's properties (wide-angle lenses send rays out in large angles, and telephoto lenses send rays in a generally forward direction). This ray intersects with the white wall. The value of this part of the wall's 'illumination' is mid-grey, and so that is the colour of the first pixel. The second pixel in the first row is then traced. Again a ray is cast, which hits the white wall and returns a grey value for the second pixel. The rest of the top row of pixels are evaluated in the same way, all returning the same grey value. After evaluating the top row, the next rows are evaluated by the same ray-tracing method, left-to-right, top-to-bottom. On the first pixel of the forth row, instead of the ray intersecting with the wall, the ray does not intersect with any object. This returns nothing as a value, and so the pixel is rendered black. The image on the left shows the result of this process when completed. You can see where the ray has hit the sphere the pixels are red, and the various grey and blue of the checkered floor where the traced ray intersects it.

The ray-tracing procedure subdivides the scene into individual pixels

Okay. So to produce an image we send rays to "scan" a scene and report back on the illumination; the amount of light at that point in the scene. This "report" uses RealSoft3D's incredible (both in versatility and complexity!) channel system.

An Explanation of Channels

Every object in your scene, whether geometry, lights or NURBS curves, has a collection of properties. The most obvious is colour, and there are others like transparency and reflectivity. This information is stored as numerical data in channels. Channels may be a misleading name; you can consider them as just database fields in a record. If you open up the properties for an object and look under the "Col" tab, you will see at the top the surface properties as defined for that object. The default attribute is Color. For the red sphere the value in this field was set to (1,0,0). We can therefore say that the sphere's Color channel is (1,0,0). If you click the attribute selector you can see a long list of different properties, all of them referring to different channels. The Transparency channel for the sphere has a value of (0,0,0) - there is no transparency. The Illumination channel also has a value of (0,0,0). This means that the amount of light originating from the sphere is nothing.

Evaluating Surfaces

When we raytrace the image, we use the information in the channels to build up the picture. Most importantly, we use the value of the Illumination channel as this tells us how much light is coming from the surface (whether this light be reflected from another light source or being radiated). The 10 by 10 pixel image shows the surface's Illumination channel at that point. But if the Illumination channel for the sphere is (0,0,0), how come it's rendered red? Which brings us to the next important point!

Changing Channel Values

A fundamental aspect of channels is that their information is not static. The channel values are initialised at the beginning of rendering but may change either as a result of RealSoft3D's default rendering calculations or as a result of a VSL material that modifies this data. The sphere's Illumination channel is initialised to (0,0,0). But when the camera ray is cast and terminates at the sphere, another step is entered into. This step evaluates the effect of light sources. It calculates the angle that rays of light from the light source strike the surface of the sphere. It uses this angle to calculate how much light would be reflected back towards the camera. And finally, it multiplies this amount of light by the Color of the surface being evaluated. Here's the obligatory example. In this diagram three light rays are marked, A, B and C.  The single lightsource is of brightness 1.0 and colour (1,1,1). This information, like all RealSoft3D data, is stored in channels but I'll come back to this later.  Suffice for now, the value of the light reaching the sphere is white, (1,1,1).

At point A, 100% of the light is reflected back to the camera.  That means 100% of (1,1,1) which of course is (1,1,1).  This is the value of the light's Illumination. Multiplying this by the surface's Color (1,0,0) gives: (1,0,0).  This is the value of the surface illumination at this point and this is added to the surface's Illumination channel.  For point B, about 50% of the light is reflected from the surface towards the camera.  50% of (1,1,1) is (0.5,0.5,0.5) which is multiplied by the surface's Colour channel.  The result added to the surface's Illumination channel is (0.5,0,0).  Finally, for point C, none of the light is reflected back to the camera.  So the surface's Color channel is multiplied by (0,0,0) to give (0,0,0), added to the surface's Ilumination channel.

An important point to realise here that there exist a record of channel values for each and every ray cast. It is easy to get confused into thinking that the Color channel of an object is shared by the whole of that object's surface; that there is one instance of the Color channel.  The properties of the object specify default values that will be copied into the relevant channels when a ray is traced.

Post Processing

This evaluation of the scene by tracing rays goes through one last step to produce the final image - the Post Processing stage. Although post-processing suggest manipulation of the data, the first action of post-processing is to turn the channel data into image data. Before rendering a scene in RealSoft3D you need to have in effect a Post Image. There is always one active, usually the default. The effects of the post-action are applied at the end of a traced ray, where it's first action is to copy the value from the surface's Illumination channel to the image's Color channel. After this, further items within the Post Image effect adjust the value in the image's Color such as adjusting the brightness, adding or changing values due to the presence of particle effects, blurring the Color channel, etc.  The value of this channel gives the colour of the pixel in the final image.

Overview

Before going a little more in-depth into the process, let's end this part with an overview of what's been said so far. Consider the rendering of the fourth line down in the 10 by 10 pixel image raytraced.

1. For the first pixel, a ray is cast. This ray does not intersect with any geometry, so the value in the surface's Illumination channel is left at (0,0,0). This value is copied into the image's Color channel for that pixel.
2. The next ray is cast which intersects with the wall. The Color channel of the wall is (1,1,1), and 48.4% of the light from the light source is reflected back into the camera lense. Therefore, 48.4% of Color is added to the surface's Illumination channel, (0.484,0.484,0.484). This value is copied to the final image's Color channel.
3. The third pixel is evaluated to the same result as the second pixel.  The Illumination channel for the image is also (0.484,0.484,0.484).
4. Same as the third pixel, the Illumination channel for this pixel is (0.484,0.484,0.484).
5. The ray cast for the fifth pixel intersects with the red sphere, Color (1,0,0). 87.1% of the light is reflected back towards the camera and the Illumination channel for this surface has (0.871,0,0) added to it. At the post-image stage, the Color channel is a copy of this.
6. The sixth ray hits the sphere where the reflected light is 71%.  The Illumination channel has added to it (0.71,0,0), and the image's Color channel is a copy of this.
7. The seventh ray finds the sphere reflecting 45.3% of the light, giving a final image Color channel value of (0.453,0,0)
8. The ray cast for the eigth pixel of this row finds the sphere in it's own shadow.  No light is reaching this part of the surface, so the Illumination channel remains at (0,0,0) and the final image Color channel copies this value.
9. The last two rays cast for the ninth and tenth pixels terminate at the wall.  They are evaluated to the same value as the other wall pixels and the Color channel for the image for these two pixels is (0.484,0.484,0.484)

At the end of tracing this row, the values in the Color channel for the image are:

Pixel	1	2	3	4	5	6	7	8	9	10
Red Component	0	0.484	0.484	0.484	0.871	0.71	0.453	0	0.484	0.484
Green Component	0	0.484	0.484	0.484	0	0	0	0	0.484	0.484
Blue Component	0	0.484	0.484	0.484	0	0	0	0	0.484	0.484

In the compilation of the final image, the colour of the pixels is calculated from the image Color channel values. A value of 1.0 means 100% which, in a 24 bit colour image, means a value of 255. The value of the Color channel is therefore multiplied by 255 to get the resultant pixel colour. For pixel 0, the resultant pixel colour components are all 0.0*255 = (0,0,0).  For pixel 1, they are all 0.484*255 = (123,123,123).  For pixel 5, the red component is 0.871*255 = 222.  Green and Blue components are both 0, so the resultant pixel colour is (222,0,0)

This conversion from image Color channel data to output image colour values occurs when the output of the rendering is a 24bit colour image. RealSoft3D can output to greater colour depths in which case the values of the Color channel are multiplied by the 100% value for the used colour depth. The final order of events is:

Ray is cast

Ray terminates when hits a surface.  Surface properties are initialised

Lighting calculations add degree of illumination to surface's Illumination channel

Post-effect entered into.  Surface's Illumination channel copied to image's Color channel

Final image.  Pixel colour derived from image's Illumination channel

Summary

RealSoft3D stores surface properties in channels. When rendering an image, rays are cast into a scene, one ray for each pixel.  When these rays hit a surface, calculations compute the amount of light reflected back to the camera as a product of light intensity and surface colour, and this is added to the surface's Illumination channel. The value in this Illumination channel is copied to the image's Color channel in the beginning of a post image effect. Finally, the value in the image's Color channel is used to calculate the final colour values of the rendered image.

Part 2 of "The RealSoft3D ray-tracing process" will look in more depth at the individual steps of RealSoft3D's rendering process and introduce shaders. Click HERE to continute to Part 2.