3 replies to this topic

Posted 13 November 2012 - 05:02 AM

Hi,

I'm working on a spotlight shader and am stumped on the specular component calculation, most specular calculations assume an angle of reflectance from the lights position against the normal of the object in question (the light being assumed to be omini in this instance).

Here is a standard omni representation...

[source lang="cpp"] float3 directionToLight= normalize(lightPosition - In.WorldPosition) float3 norm = normalize( reflect( -directionToLight, normalize(In.WorldNormal) ) ); // Calculate the half angle: the half the angle between our light direction and camera view float3 halfAngle = normalize( In.CameraView - directionToLight); //Get Specular component specular = pow( saturate( dot( norm, halfAngle ) ), specularPower ) * specularLightColor * specularIntensity;[/source]

The Spotlight is a different beast as it has direction as well as position, and this calculation is not physically correct as the light source could be rotated away from from the object and the calculation would yield a result for a positive specular value. Does anyone have any good examples of the correct specular calculation?

Thanks in advance.

I'm working on a spotlight shader and am stumped on the specular component calculation, most specular calculations assume an angle of reflectance from the lights position against the normal of the object in question (the light being assumed to be omini in this instance).

Here is a standard omni representation...

[source lang="cpp"] float3 directionToLight= normalize(lightPosition - In.WorldPosition) float3 norm = normalize( reflect( -directionToLight, normalize(In.WorldNormal) ) ); // Calculate the half angle: the half the angle between our light direction and camera view float3 halfAngle = normalize( In.CameraView - directionToLight); //Get Specular component specular = pow( saturate( dot( norm, halfAngle ) ), specularPower ) * specularLightColor * specularIntensity;[/source]

The Spotlight is a different beast as it has direction as well as position, and this calculation is not physically correct as the light source could be rotated away from from the object and the calculation would yield a result for a positive specular value. Does anyone have any good examples of the correct specular calculation?

Thanks in advance.

Posted 13 November 2012 - 02:25 PM

As far as I know, you can use the same formulas for specular component as you use for point light sources. The direction and the inner and outer cone values affect only the attenuation, which is applied to the specular component as well. This attenuation prevents the spot light to affect objects outside of the illuminated zone.

Cheers!

[edit]

http://content.gpwiki.org/index.php/D3DBook:(Lighting)_Direct_Light_Sources some lighting information here, although the spot light code doesn't seem to implement specular term, it shows how to calculate the attenuation, which can be applied to your specular factor.

Cheers!

[edit]

http://content.gpwiki.org/index.php/D3DBook:(Lighting)_Direct_Light_Sources some lighting information here, although the spot light code doesn't seem to implement specular term, it shows how to calculate the attenuation, which can be applied to your specular factor.

**Edited by kauna, 13 November 2012 - 02:30 PM.**

Posted 13 November 2012 - 04:03 PM

Your specular calculations are really weird, you seem to be combining bits from Phong and Blinn-Phong. Blinn-Phong uses N dot H, where N is the surface normal. You're using the direction of the light reflected about the surface normal, which is incorrect. You'll also want to multiply specular by N dot L, since this is implicit in all BRDF's and also prevents a few artifacts that can occur. Here's a revised version of your code:

To make that look even better you can add the normalization term to your specular.

As for spot lights, the only difference between a point light and a spot light is an additional attenuation term. Typically for a point light you use some attenuation term that you multiply into the light intensity that's based on the distance from the surface to the light source. For a spot light you can use this same distance attenutation, and then multiply it with an additional term based on the light direction and cone angles.

float3 directionToLight = normalize(lightPosition - In.WorldPosition) float3 norm = normalize(In.WorldNormal); float nDotL = saturate(dot(directionToLight, norm)); // Calculate the half angle: the half the angle between our light direction and camera view // Note: CameraView should be equal to (cameraPosition - worldPosition) float3 halfAngle = normalize(normalize(in.CameraView) + directionToLight); //Get Specular component specular = pow(saturate(dot(norm, halfAngle)), specularPower) * specularLightColor * specularIntensity * nDotL;

To make that look even better you can add the normalization term to your specular.

As for spot lights, the only difference between a point light and a spot light is an additional attenuation term. Typically for a point light you use some attenuation term that you multiply into the light intensity that's based on the distance from the surface to the light source. For a spot light you can use this same distance attenutation, and then multiply it with an additional term based on the light direction and cone angles.

**Edited by MJP, 13 November 2012 - 04:04 PM.**

Posted 14 November 2012 - 12:16 PM

I would have thought that the specular component was affected by the rotation of the light, but I tried an experiment in 3D Studio Max to sweep (via a rotate) a spotlight across a plane and the result was that the specular highlight stayed in the same place. If Autodesk can do it, so can I!

The main result of this is that I don't have to do anything different for my spotlight, except attenuate the specular component by the spotlight calculation.

The secondary result of this is that I feel like an amateur for posting 'really weird' code, anyway thank you MJP for the pointer to my incorrect specular calculation, I have now 'de-weirded' it.

The main result of this is that I don't have to do anything different for my spotlight, except attenuate the specular component by the spotlight calculation.

The secondary result of this is that I feel like an amateur for posting 'really weird' code, anyway thank you MJP for the pointer to my incorrect specular calculation, I have now 'de-weirded' it.