So in reality, everything's a light source. Yes, the sun is a light source. But the sky is also a light source, filling in areas not illuminated by direct sunlight with a soft blue glow. Your pants are a light source, lighting up the ground in front of you with pants-colored light.

Does this matter? Well, does this look like OpenGL?

This is an image computed using "radiosity".

- Power and Energy carried by light (e.g., power received at a 100% efficient photocell)

- It's not energy, it's energy per unit time.

- Measured in Watts

- Watt=1 Joule/second, a small unit of power

- =1 Newton * meter/second (a quarter-pound weight lifted at one yard per second)
- =1 Volt * Ampere (or 0.01 amps at 100 volts)

- =1/746 horsepower
- =1/4.186 calories (calorie = 1cc of water heated by one degree C per second; food is measured in *kilocalories*, but they leave off the "kilo")
- Also measured in "lumens"-- power as perceived by the human eye. At a wavelength of 555nm, one lumen is 1/680 watt.

- AKA flux, radiant flux
- Irradiance: light power per unit surface area leaving or arriving at a surface

- Measured in watts per square meter [of receiver or transmitter area].
- So multiply by the receiver's area to get watts. (Assuming everything's facing dead-on.)

- E.g., the Sun delivers an irradiance of 700W/m
^{2}to the surface of the Earth.

- So a 10 square meter solar panel facing the sun would receive 7000W of power.
- So a 1.0e-10 square meter CCD pixel open for 1.0e-3 seconds would receive 7.0e-11 joules of light energy, or a couple hundred million photons (visible light photons have energy of about 4e-19 Joules).
- Irradiance as perceived by the human eye can be measured in "lux" (lumens per square meter).

- Radiance: "brightness" as seen by a camera or your eye.
- Measured in (get this) watts per square meter [of receiver
area] per steradian [of source coverage, as measured from the
receiver], or just W/m
^{2}/sr. (See below for details on steradians) - E.g.,
from the Earth the Sun is about 1 degree across, or about 1/60 radian,
so assuming it's a tiny square it would cover 1/3600 steradians.
The radiance of the sun is thus about 700W/m
^{2}per 1/3600 steradians, or 2.3 megawatts per square meter per steradian (2.3 MW/m^{2}/sr). - This
means if a lens sets it up so from one spot you see the sun at a
coverage of 1 steradian, that spot receives an irradiance of 2.3MW/m
^{2}of power! - Radiance is unchanged through free space.
- That is, Radiance does *not* depend directly on distance: if the source coverage is unchanged, the radiance is unchanged. That is, the pixel doesn't get brighter just because the geometry is right in front of it!
- Suprisingly, Radiance is even unchanged when passing through
a *lens*. Sunlight focused through a lens has the same radiance,
but because the lens changes the Sun's *angle* (and hence solid angle
coverage) of the light, the received irradiance is bigger.

- Solid angle: "bigness" of light source as seen by a receiver.
- Measured in "Steradians" (abbreviated "sr").
- A
light source's solid angle in steradians is defined as the ratio of the
area of the light source when projected onto the surface
of a sphere centered on the receiver. This area is then divided
by the sphere's radius
squared, or equivalently, is always performed with a unit-radius sphere.

- See entries in Mathworld or the figure in the Lighting Design Glossary.

- Examples:

- The maximum coverage is the whole sphere, which has area 4 pi * radius squared, so the maximum coverage is 4 pi steradians.
- A hemisphere is half the sphere, so the coverage is 2 pi steradians.
- A small square that measures T radians across covers approximately T
^{2}steradians. This expression is exact for an infinitesimal square.

- A flat receiver surface only receives part of the light from a
light source low on its horizon--this is Lambert's cosine illumination
law. Hence for a flat receiver, we've got to weight the incoming solid angle by a
factor of "cosine(angle to normal)".

- For a source that lies directly along the receiver's normal, this doesn't affect the solid angle--it's a scaling by 1.0.
- For a source that lies right at the receiver's horizon, this factor totally eliminates the source's contribution--it's a scaling by 0.0.
- For a hemisphere, this cosine factor varies across the surface, but it integrates out to a weighted solid angle of just 1 pi steradians (the unweighted solid angle of a hemisphere is 2 pi steradians).
- You only want the solid angle for computing illumination from a polygon to a point. For
illumination between two polygons, you actually want to compute the
"form factor" between the polygons, which you can either approximate
using the solid angle, or compute exactly using Schröder and Hanrahan's 1990 paper.

To do global illumination, we start with the radiance of each light source (in W/m

To do global illumination for diffuse surfaces, then, for each "patch" of geometry we:

- Compute how much light comes in. For each light source:
- Compute the cosine-weighted solid angle for that light source.
- This depends on the light source's shape and position.
- It also depends on what's in the way: what shapes occlude/shadow the light.

- Compute the radiance of the light source in the patch's direction.
- This might just be a fixed brightness, or might depend on view angle or location.

- Multiply the two: incoming irradiance = radiance times weighted solid angle.

- Compute how much light goes out. For a diffuse patch,
outgoing radiance is just the total incoming irradiance (from all light
sources) divided by pi.

The oldest, best-known, and least accurate way to compute a solid angle is to assume the light source is small and/or far away. If either of these are true, then the light source projects to a little dot on the hemisphere, and we just need to figure out how the area (and hence solid angle) of that dot scales with distance. Well, the height of the dot scales like 1/z, and the width of the dot scales like 1/z, so together the area (width times height) scales like 1/z

For example, say the light source is an isotropic 1-meter radius disk, sitting z meters away and directly facing you. What's the falloff with z? It turns out to be 1/(z

vec3 light_vector=vec3(0.0);

for (int i=0;i<light.size();i++) { // Loop over vertices of light source

vec3 Ri=normalize(receiver-light[i]); // Points from light source vertex to receiver

vec3 Rp=normalize(receiver-light[i+1]); // Points from light source's *next* vertex to receiver

light_vector += acos(dot(Ri,Rp)) * normalize(cross(Ri,Rp));

}

float solid_angle = dot(receiver_normal,light_vector);

You can actually just dump this whole thing directly into GLSL, and it will run. Two years ago, the "acos" kicked it out into software rendering, which was 1000x slower. Today, on the *same* hardware, the *same* GLSL program runs in graphics hardware, and is pretty dang fast too--a sincere thank you goes out to the ATI GLSL driver writers!

You can speed it up a bit (and mess it up too at close range!) by first noting that

` light_vector += acos(dot(Ri,Rp)) * normalize(cross(Ri,Rp))`

is an angle-weighted, normalized cross product. In theory we could compute the angle weighting just as well with

` light_vector += asin(length(cross(Ri,Rp))) * normalize(cross(Ri,Rp))`

This isn't actually useful in practice, because it breaks down for
nearby receivers--the angle between Ri and Rp exceeds 90 degrees, and
asin starts giving the wrong answer, while acos keeps working all the
way out to 180.However, for distant receivers, which have small angles between Ri and Rp, we're close to having asin x == x. Thus the above is very nearly equal to:

` light_vector += length(cross(Ri,Rp))) * normalize(cross(Ri,Rp))`

Now, the length of a vector times the direction of the vector gives the
original vector, so we can actually just sum up the cross products:light_vector+=cross(Ri,Rp);In practice, this is very difficult to distinguish visually from the real equation, except at extreme close range where this approximation comes out a bit darker than the real thing.

It is actually possible to clip away the occluded parts of the light source, and then compute the solid angle of the non-occluded pieces. I've done this in software. It's actually tolerably fast as long as you're willing to limit yourself to a few dozen polygons per scene. But with any realistic number of polygons, computing the exact solid angle becomes very painfully slow. This is where approximations come in, which we'll cover next class.