A Few Updates

Well ive been working on a few updates, had a crappy few months due the death of my grandfather. But well im back now. I have updated the camera class to use Quaternion, and i have also fixed some hdr lighting issues and made the atmopshere look more realistic. I am currently working on a ocean shader, but thats a bit a fail upwards senario. Anyways heres some images of what i have been working on lately. Its a littlebit over exposed in some places but it looks nice :).

More of an update

Well again i thought that this deserves an update, still havent fixed that normal problem , but here it is. I managed to implement a almost identical set of function like libnoise, using libnoise.net as a ref you can find it over at http://libnoisedotnet.codeplex.com/. It works great beacuse i can use it to generate a planet heightmap which warps perfectly around my planet sphere, and look's a hell of a lot better then a few ocatves of fbm. Here are some screenshots of my current progress and also a work in progress shot of me failing to implement slope based terrain texturing (cant win the all in 1 day :) ).

Slope based terrain texturing

And here they are my latest procedural heightmap functions at work, using a bunch of random seed's.

Oh My Days ive finally got it!!

Well this is more of a progress update then a post, i have finally managed to get the cracks fixed perfectly, still need to fix the normal generation but im getting there. Also i have improved the fps alot i get an avarage of 80fps and even 120fps at some points. I realised that i was not generating the patch indexes properly, which was causing them to point to a completely wrong vertex. Ill give ya a few screenshots but nothing major to see just you wont notice any cracks any more. If anybodys got a any good ideas on hiding the lod transition, with the normal calculation let me know at the moment im looking on generating normal maps on a per patch basis.

Just a little eye canddy

I decided to post a screeny i am working on some new noise functions, so far i generate a heightmap texture using a mixture of fbm, and simplex noise. I will soon be working on noise features, craters volcanoes, etc, just thought it looked cool sorry guyz no technical stuff today. I set the R texture channel to blue beacuse it think it looks cool, with my fake water :).

Creating Normal Maps on the cpu

I have come across this nice bit of code on gamedev.net forums. Which i think will come in handy for a loads of people as i cant seem to find alot of examples of generating normal maps from heightmaps.

At the moment Indevilage Engine (yes my engine is called Indevilage) uses normals generated on a per patch basis but eventually i want to use high res normal maps generated on the gpu. Heres the function and an example of what it does.

I know this is a bit of a side step from my normal posts but i haven really posted anything usefull i thin its time i started.

1:   public static Texture2D NormalMapFromHeightMap(GraphicsDevice Device,Texture2D heightmap)
  
2:        {
  
3:          int size = heightmap.Width;
  
4:          Color[] colorArray = new Color[size * size];
  
5:          heightmap.GetData<Color>(colorArray); 
  
6:          Color[] normalArray = new Color[size * size];
  
7:          Vector3 xyzdist = new Vector3(0.01f, 0.01f, 0.01f);//new Vector3(1.0f, 1.0f, 1.0f); 
  
8:          for (int i = 0; i < size - 1; i++) 
  
9:          { 
  
10:            for (int j = 0; j < size - 1; j++) 
  
11:            { 
  
12:              float sample1 = (colorArray[i + (j + 1) * size].R + colorArray[i + (j + 1) * size].G + colorArray[i + (j + 1) * size].B); 
  
13:              float sample2 = (colorArray[i + j * size].R + colorArray[i + j * size].G + colorArray[i + j * size].B); 
  
14:              float sample3 = (colorArray[(i + 1) + j * size].R + colorArray[(i + 1) + j * size].G + colorArray[(i + 1) + j * size].B); 
  
15:              Vector3 surfacesample0 = new Vector3(i * xyzdist.X, j * xyzdist.Y, sample1 * xyzdist.Z); 
  
16:              Vector3 surfacesample1 = new Vector3(i * xyzdist.X, (j + 1) * xyzdist.Y, sample2 * xyzdist.Z); 
  
17:              Vector3 surfacesample2 = new Vector3((i + 1) * xyzdist.X, j * xyzdist.Y, sample3 * xyzdist.Z); 
  
18:              Vector3 surfacevec0 = surfacesample1 - surfacesample0; 
  
19:              Vector3 surfacevec1 = surfacesample2 - surfacesample0; 
  
20:              surfacevec0.Normalize(); 
  
21:              surfacevec1.Normalize(); 
  
22:              Vector3 surfacenormal = new Vector3(); 
  
23:              Vector3.Cross(ref surfacevec0, ref surfacevec1, out surfacenormal); 
  
24:              surfacenormal.Normalize();
  
25:              surfacenormal.Z = -surfacenormal.Z;
  
26:              surfacenormal.X = (surfacenormal.X + 1.0f) / 2.0f;
  
27:              surfacenormal.Y = (surfacenormal.Y + 1.0f) / 2.0f;
  
28:              surfacenormal.Z = (surfacenormal.Z + 1.0f) / 2.0f;
  
29:              normalArray[i + j * size] = new Color(surfacenormal);
  
30:            } 
  
31:          } 
  
32:          Texture2D texture = new Texture2D(Device, size, size); 
  
33:          texture.SetData(normalArray);
  
34:          return texture; 
  
35:        }  

Heightmap

NormalMap

I havent actually had time to test this in my engine, but it seems to generate the right values, i hope this is of help to anybody.

Using a logarithmic z buffer

Well i have decided to enable the crack fixing still not perfect but it works, the only problem is with such a large object you get shimering due to zbuffer fighting, the way this is solved by most people is to use a logarithmic z buffer, the way to do this is to calculate the depth in the shader. It was actually very simple to do, just add a few lines of code and there we have it.

1:       const float Cc = 1.0; 
  
2:       const float Far = 1000000000.0; 
  
3:       output.Position.z = log(Cc*output.Position.z + 1) / log(Cc*Far + 1) * output.Position.w;   

I show how this works in the following screenshots.

without the logarithmic z buffer

with logarithmic z buffer

It has now brought to my attention a few other things i can see but for now this will do :).

BoundingBox's A pain in the but

I have spent a long time trying to ificiantly calculate bounding boxes for my terrain patches, i have resorted to bruteforce once the patch is created, the code is just a modifyed version of the standard xna method for calculting mesh bounding boxes. Here is a little update pic and the code used. I dont think this really needs an explanation, it just checks all vertices and then calculates the max and min bounds.

  public BoundingBox CalculateBoundingBox()
  
  {
  
    Vector3 Max = new Vector3(float.MinValue, float.MinValue, float.MinValue);
  
    Vector3 Min = new Vector3(float.MaxValue, float.MaxValue, float.MaxValue);
  
     for (int i = 0; i < VertexData.Length; i ++)
  
     {
  
       Vector3 vectCurrentVertex = VertexData[i].Position;
  
       if (VertexData[i].Position.X < Min.X) Min.X = VertexData[i].Position.X;
  
       if (VertexData[i].Position.X > Max.X) Max.X = VertexData[i].Position.X;
  
       if (VertexData[i].Position.Y < Min.Y) Min.Y = VertexData[i].Position.Y;

  
       if (VertexData[i].Position.Y > Max.Y) Max.Y = VertexData[i].Position.Y;
  
       if (VertexData[i].Position.Z < Min.Z) Min.Z = VertexData[i].Position.Z;
  
       if (VertexData[i].Position.Z > Max.Z) Max.Z = VertexData[i].Position.Z;
  
     }
  
     return new BoundingBox(Min, Max); 
  
  }