Friday, 27 January 2012

Normal Maps from Photos

Normal mapping is a vital part of 3D modelling and rendering these days. In some cases it take hours off your modelling time, and in others hours off your rendering time (and sometimes both).

A while ago I came across this tutorial about making normal maps from photos. I finally got round to trying it.

Introducing the fruits of my labours - this example of mine is the side of a well worn pound coin.

You don't need Photoshop to do this, and to prove it I used the GIMP image manipulation program to make the normal map, and then applied and rendered in Carrara below.

You can see from this render that it works, but some of the edges need to be sharper.

A brief critique of this attempt shows the angle of the light source was too low when photographing the coin, making long shadows, and positioning the light source slightly higher would shorten the shadows and give sharper less buttery results, but having tested it using the 3D preview mode in the GIMP Normal Map Filter, it isn't half bad, and much quicker than trying to build a high res mesh, or digitally painting a height map for conversion to a normal map using tools like the Nvidia Normal Map Filter.

For some objects you really should give this a try.

Of course, even this early attempt is usable if supported by other channels such as shine and highlight (specular) using a grayscale map based on the same photos as the normal map. I also used the same grayscale map to mix two shades of coin colour to show a diffrence where dirt gets ingrained and where it gets rubbed off and used the same map again to give two levels of reflection because dirt is not as reflective as metal. Combining all these with the normal map brings a whole different level of realism.

I'll explain how I used the same photos as the normal map to make the grayscale specular map that I also used for reflection and grime in a future post.