Transcript Rendering Algorithms
Shading, Surfaces & Textures
Outline
Polygon shading models Constant, interpolated, polygon mesh, Lambert, Gouraud, Phong Lighting models Lambert, Phong, Blinn Surfaces properties Surface mapping Texture Bump Displacement …
Lighting algorithms
Most algorithms use the polygon or surface
normal
(vector that is perpendicular to the polygon) The amount of light reflected depends on the angle of the incident light, e.g. for diffuse reflection:
L N
Constant shading
Also known as
Lambert
,
faceted
and
flat shading
Calculates a single intensity value for each polygon Valid if following assumptions are true: Light source is at infinity (i.e. angle between light and surface is constant) Viewer is at infinity (i.e. angle between viewer and surface is constant) Polygon is not an approximation of a curved surface
Interpolated shading
Shading is linearly interpolated across polygon Faster than calculating each pixel, better than flat shading Still assumes that polygon represents the true surface
Polygon mesh shading
Many sets of polygons actually are approximations to curved surfaces Shading the polygons to realistically simulate the surface is therefore important Several methods exist to achieve this, the most important of which are
Gouraud
and
Phong
shading.
Gouraud shading
(Henri Gouraud, PhD 1971)
This is a form of
intensity interpolation
shading Starts from knowing the normal of the surface at each vertex of the polygon These may be stored with the mesh or may be averaged from the adjacent polygons Next step is to calculate the
vertex intensities
This will use the vertex normal and the illumination algorithm
Gouraud shading (2)
The polygon is then shaded by: Interpolating the intensities along each edge Interpolating across the polygon between the edges along each scan line
I 2 I P I 1 I 3
Gouraud algorithm
Compute S A , S B , S C S i = shade of point i.
for triangle ABC.
For a scanline XY, compute S X , S Y interpolating e.g. t AB = |AX| / |AB|.
S A = t AB * S A + (1-t AB )*S B X Compute S P By interpolating A S X P by S P B between S X and S Y S Y Y C scanline
Gouraud example
Phong shading
(Bui-Tong Phong, PhD 1975)
This is a form of
normal-vector interpolation
shading Starts from the normal vector at each vertex again, but interpolates the vector rather than the intensity
Phong shading (2)
In general, this yields much more realistic results, especially for specular surfaces Gouraud shading may miss specular highlights It is several orders of magnitude slower than Gouraud
Problems with interpolated shading
Polygonal silhouette Perspective distortion Orientation dependence Shared vertices Unrepresentative vertex normals
Polygonal silhouette
Approximation is polygonal, so no matter how good the shading algorithm the silhouette will be faceted Can be improved by increasing the polygon count at expense of processing time
Perspective distortion
Because the interpolation is based on the scan lines, the number of interpolation increments per polygon depends on the angle of the polygon to the viewer Steep angle = few increments Shallow angle = more increments
Orientation dependence
Because interpolation occurs between vertices and along scan lines, result depends on orientation of polygon relative to scan lines
Shared vertices
Problems occur when adjacent polygons do not share vertices This can lead to discontinuities in shading
Unrepresentative vertex normals
Vertices may not adequately represent the surface geometry Usually can be solved by further subdividing surface
Surfaces
Definition of visual characteristics of a surface Often defined by a ‘shader’ or ‘material’ Specifies colour, how shiny it is, etc.
Surface parameters
Colour usually defined by RGB values or via GUI Ambient is a false parameter that defines how ‘ambient’ light is treated Diffuse parameter specifies how much light in total is evenly reflected Specularity defines how shiny a surface is (a high value will have highlights like a billiard ball) May also have controls for highlight size and colour
Advanced surface parameters
Reflectivity defines how much of the surrounding environment is reflected by the surface (like a mirror) Transparency defines of the background is visible through the surface Translucency defines how much light is transmitted through the surface Refractivity defines how much light is bent as it enters and leaves the material
Surfaces in Maya
Basic materials: Lambert Phong Phong E Blinn Special materials: Anisotropic, Layered Shader, Ocean Shader, Ramp Shader, Shading Map, Surface Shader, Use Background
Lambert
This is the simplest material It creates a matt surface with diffuse and ambient components but no highlights The shading interpolates between adjacent polygon normals so the surface appears smooth It uses
Lambert’s cosine law
Lambert’s Cosine Law
“The reflected luminous intensity in any direction from a perfectly diffusing surface varies as the cosine of the angle between the direction of incident light and the normal vector of the surface”
Johann Lambert, 1728-1777.
Lambert’s Cosine Law
Ideal diffuse surfaces obey this cosine law
I d
Often called
Lambertian
=
k d I incident
cos =
k d I incident
(
N
·
L
).
surfaces.
where
k d
is the diffuse reflectance of the material.
Wavelength dependent, so usually specified as a colour.
Maya Lambert properties
Phong
(Bui-Tong Phong, 1975)
This lighting model includes
specularity
This takes into account that the amount of light you see depends on the viewer’s angle to the surface as well as the light’s angle His original formula for the specular term: W(i)(cos s ) n s is the angle between the view and specular reflection directions.
“W(i) is a function which gives the ratio of the specular reflected light and the incident light as a function of the the incident angle i.” Ranges from 10 to 80 percent.
“n is a power which models the specular reflected light for each material.” Ranges from 1 to 10.
Phong lighting model
More recent formulations are slightly different Replace
W
(
i
) with a constant
k s
, independent of the incident direction
I s
=
k s I incident
=
k s I incident
V
( cos
V
·
n R
)
n .
is the view direction.
R
is the specular reflection direction L N R V
Variation of
n
Maya Phong properties
Maya’s ‘Phong E’
Uses a simplified model, faster to render than pure Phong
Maya’s Blinn
(Jim Blinn, 1977)
Based on
Blinn-Phong
shading (an adaptation of Phong) Offers great control over specularity but in general will not look as ‘shiny’ as a Phong material
II
Maya Blinn properties
2D texture mapping
A ‘cheap’ way of enhancing the the surface definition So far, surfaces have been defined with a plain colour In real life, many surfaces are multi-coloured E.g. Wood looks ‘wooden’ because of many different colour variations in the grain
2D texture mapping (2)
2D texture mapping involves applying a 2D image to the surface of a 3D object In this case the term ‘texture’ applies to the image rather than the more traditional ‘feel’ of the surface Textures may be ‘real’ (e.g. scanned), manually generated (e.g. with a paint program) or procedurally generated (e.g. by a computer program)
2D texture mapping (3)
The process of applying the texture is the ‘mapping’ The two mapping processes: Projecting the image onto the surface Stretching the image to fit on the surface
Projection mapping
In projection mapping the image is projected through space Wherever it ‘hits’ the surface the surface becomes the colour of the texture The purest form of this is
planar projection
Cylindrical projection
Image is bent into a cylinder before projection takes place Object is then placed ‘in’ the cylinder
Spherical projection
Image bent onto an imaginary sphere before projection Object placed ‘in’ sphere
Limitations
Unless mapping on a perfect flat surface, cylinder or sphere, all projection methods will have a tendency to ‘streak’ Parameterised texture mapping offers a solution to this problem
Parameterised (UV) mapping
Imagine printing the texture image onto thin transparent film Then stretch the film over the surface This is the principle of parameterised texture mapping
Parameterised mapping (2)
The 2D texture image is (obviously) an array of rectangular pixels These can be referenced by Cartesian coordinates, usually with the lower left pixel being (0,0) (and for a 512 x 512 image the top right being (511,511)) Each pixel rectangle is then mapped to a corresponding area of the 3D surface
Parameterised mapping (3)
You therefore have to divide the surface into the same number of steps The surface patch is defined to use the coordinate system
U
locations on it and
V
to specify The area defined by (0,0) on the surface has the colour applied to it from the image pixel at (0,0)
Procedural textures
Textures can be generated by
procedures
rather than using predefined images This is good for regular patterns (e.g. brickwork) or (semi-)random patterns (e.g. noise)
Maya procedural textures
Got to here!
Other types of mapping
So far, all examples have used the colour of the image to affect the colour of the surface - this is
colour texture mapping
Other options are
transparency mapping
,
bump mapping
and
displacement mapping
Transparency mapping
Use pixel value to define transparency of object Usually uses grey values, white = opaque & black = transparent
Bump mapping
Uses pixel values to ‘perturb’ the surface normals Used to model tiny variations in surface contour Actual surface doesn’t change geometry, but appears to because normal is shifted Good for ‘orange peel’ like effects, but outline is still smooth
Displacement mapping
Uses the intensity of the pixel to actually shift part of the surface up (for white) or down (for black) Used to model surface irregularity, but unlike bump mapping the surface geometry actually changes
Other image-based mapping
Incandescence mapping Pixel values = ‘glow brightness’ Specularity mapping Pixel values = specularity Reflectivity mapping Pixel values = reflectivity
3D texture mapping
Some of these effects can be achieved by ray tracing, but this is expensive Often, realistic objects require more than just 2D images mapped onto them E.g. Shiny surfaces (such as chrome) should reflect the environment that surrounds them We can use
environmental reflection mapping
(sometimes called just
reflection mapping
)
Reflection mapping
Cubic (most common) Six views of the surrounding world are generated and mapped on the object using cubic texture mapping Six images may be rendered views Spherical Use a single image that is spherically texture mapped Difficult to generate the single image correctly
Solid texture mapping
Useful for things such as wood grain, marble, etc.
Create a volume of texture Object then ‘floats’ in this volume and surface points adopt colours defined in the volume Can also be used to model transparency and other effects
Summary
Shading algorithms Flat, Gouraud, Phong Lighting algorithms Lambert, Phong… Surface properties Diffuse, specular… Surface mapping Texture, bump, displacement…