3D Denoising on BCC
Eurovis 2006 Publication
2D Reconstruction
3D Cell Projector
Maya-made Scenes
Radiance Scenes
Intermediate Textures
Subdivision Surfaces
Ray Tracer
3D Tetris
2D Tetris

Scenes from Radiance

The two scenes are based on two tutorials.

You can find the radiance home page here.

Night Day
night.jpg (9412 bytes) day.jpg (11610 bytes)
Corner Glass building outside
corner.jpg (7227 bytes) glassBuilding.jpg (5644 bytes)
Spotlight Procedural rainbow on ceiling
roomMod.jpg (11049 bytes) rainbowCeiling.jpg (24783 bytes)
Procedural Rippling Effect Extra Shininess
boxTex.jpg (17570 bytes) ripple.jpg (19549 bytes)
Fish Eye View  
fish.jpg (13462 bytes)  
Old scene 2 A simple flower I made
tut2.jpg (18503 bytes) flower.jpg (15856 bytes)
Effect of the -ab parameter  
According to the rpict manual, the parameter 'ab' (an integer), sets the number of ambient / diffuse bounces. The value of 'ab' must be 0 or greater.

Conceptually, 'ab' varies the amount of calculations done for inter-object reflections. At 'ab' = 0, inter-object reflections is not taken into account, that is why the ceiling in ab0.jpg is missing details compared to images generated using other values of 'ab'. When the value of 'ab' exceeds 0 (i.e. inter-object reflections enabled), notice there is little difference between the resulting images (observe the ceilings). This is because the most dominate inter-object contribution comes from the first inter-object reflection, which is the motivation behind whitted raytracing. Still, a higher value of 'ab' generates a more accurate rendering of the scene, and the image looks smoother as a result (note how the ceilings become smoother in its details as the value of 'ab' goes up).
ab = 0 ab = 1
ab0.jpg (6008 bytes) ab1.jpg (6487 bytes)
ab = 2 ab = 4
ab2.jpg (6472 bytes) ab4.jpg (6571 bytes)
Effect of the -pj parameter  
According to the radiance manual, 'pj' sets the pixel sample jitter, and its value must be greater than 0. The default value of 0 samples pixel centers. A value of 1.0 randomly distribute samples over full pixels.

This is my conceptual understanding of the influence of 'pj' with reference to the manual entry. When 'pj' = 0, each ray is shot from the center of pixels for sampling the scene. A value of 1.0 will result in a randomly selected location in each pixel where-through a ray is cast. If the value is between 0 and 1, i.e. 0.5, the ray is cast over a random area of 0.5*0.5=0.25, or 1/4 of a pixel around the center of the pixel. Therefore, if 'pj' > 1.0, the ray may actually be cast through other pixels and not the current one under consideration, that is why the image quality greatly deteriorates (i.e. the image becomes runny) as 'pj' becomes > 1.0. I have shown up to pj = 2.0 below, it is not difficult to imagine (nor to show) that for example, 'pj' = 10 will produce a very 'runny' rendering.

This method may work as a crude anti-aliasing technique, but it may not produce satisfactory images due to its random nature.
pj = 0.0 pj = 0.5
pj0.0.jpg (5816 bytes) pj0.5.jpg (5990 bytes)
pj = 1.0 pj = 2.0
pj1.0.jpg (6101 bytes) pj2.0.jpg (6318 bytes)
Effect of the -ds parameter  
According to the radiance manual, 'ds', sets the direct sampling ratio. That means, a light source will be subdivided until the width of each sample area divided by the distance of the light is below this ratio.

Since 'ds' has to be 0 or greater, I chose values 0.0, 0.1, 0.01 and 0.001. I expect that when 'ds' is small, the light source will be subdivided more for any given scene and a more accurate rendering will be produced. In other words, more shadow rays will be used to sample a scene as 'ds' decreases and remains positive.

I will explain this concept with respect to the series of images I produced. When 'ds' is 0, only one shadow ray is sent to each light source. That is why the circular shadow below the table has very sharp edges. This is not true in real life, and we would expect the shadow to have soft edges. So, I then tried 'ds' = 0.1, 0.01 and 0.001. As you can see, the circular shadow becomes increasingly softer when ds = 0.01 and ds = 0.001. Of course, when ds is very small, the change in the amount of details becomes invisible. So, it doesn't make sense to go any less than 0.001 for this particular view of the scene.

There is not a distinct difference between ds = 0 and ds = 0.1 because 0.1 is a fairly large ratio. Of course, anything greater than 0.1 will not make much of a difference either from this view. Please note however, that 0.1 might have produced a difference in other parts of the image that are not visible from this view.
ds = 0 ds = 0.1
ds0.000.jpg (6031 bytes) ds0.100.jpg (5994 bytes)
ds = 0.01 ds = 0.001
ds0.010.jpg (5686 bytes) ds0.001.jpg (5682 bytes)
Effect of the -dj parameter  
This parameter is somewhat similar to pj in that it modifies the direct source sample jitter, according to the rpict manual. The manual emphasizes that the value has to be between 0 and 1, otherwise, aiming failures may result.

'pj' by default is 0. That means one ray is shot through the center of each pixel. While pj modifies the pixel area through which to cast the ray, dj modifies the source sample area from which to (randomly) pick the sample. For example, when dj is large (i.e. 1.0), when the ray hits a light source, it does not directly sample that light source. Instead, it randomly samples from an area equal to dj*dj about the intersection point. It is obvious that the bigger the value of dj, the more jittering, or randomness, will be present in the rendering. These results are exemplified by my pictures below.
ds = 0 ds = 0.3
dj0.0.jpg (5703 bytes) dj0.3.jpg (6113 bytes)
ds = 0.6 ds = 1.0
dj0.6.jpg (6213 bytes) dj1.0.jpg (6988 bytes)


Tai Meng | 孟泰 | Last Updated: May 01, 2013