2D perlin noise in C

Question

When I implement it in my code (Raytracing), and apply it on a sphere, I get a uni-color sphere, with one stripe of darker pixel on it. When I change the random float generator, I got the basic linear noise, which isn't my goal. Can you explain what I missed?

Here is my code:

#include    <stdlib.h>
#include    <math.h>

float       noise(int x, int y)
{
  int       n;

  n = x + y * 57;
  n = pow((n << 13), n);
  return (1.0 - ( (n * (n * n * 15731 + 789221) + 1376312589) & 0x7fffffff) / 1073741824.0);
}

float       interpolate(float a, float b, float x)
{
  float     pi_mod;
  float     f_unk;

  pi_mod = x * 3.1415927;
  f_unk = (1 - cos(pi_mod)) * 0.5;
  return (a * (1 - f_unk) + b * x);
}

float       smooth_noise(int x, int y)
{
  float     corners;
  float     center;
  float     sides;

  corners = (noise(x - 1, y - 1) + noise(x + 1, y - 1) +
         noise(x - 1, x + 1) + noise(x + 1, y + 1)) / 16;
  sides = (noise(x - 1, y) + noise(x + 1, y) + noise(x, y - 1) +
       noise(x, y + 1)) / 8;
  center = noise(x, y) / 4;
  return (corners + sides + center);
}

float       noise_handler(float x, float y)
{
  int       int_val[2];
  float     frac_val[2];
  float     value[4];
  float     res[2];

  int_val[0] = (int)x;
  int_val[1] = (int)y;
  frac_val[0] = x - int_val[0];
  frac_val[1] = y - int_val[1];
  value[0] = smooth_noise(int_val[0], int_val[1]);
  value[1] = smooth_noise(int_val[0] + 1, int_val[1]);
  value[2] = smooth_noise(int_val[0], int_val[1] + 1);
  value[3] = smooth_noise(int_val[0] + 1, int_val[1] + 1);
  res[0] = interpolate(value[0], value[1], frac_val[0]);
  res[1] = interpolate(value[2], value[3], frac_val[0]);
  return (interpolate(res[0], res[1], frac_val[1]));
}

float       perlin_two(float x, float y)
{
  float     total;
  float     per;
  float     amp;
  int       hz;
  int       i;
  int       octave;

  total = 0.0;
  per = 0.5;
  octave = 10;
  i = 0;
  while (i < octave)
    {
      hz = pow(2, i);
      amp = pow(per, (float)i);
      total += noise_handler(x * (float)hz, y * (float)hz) * amp;
      i += 1;
    }
  return (total);
}

EDIT: I spot an error in the Noise function (I consider the XOR operand like a power function ... Now I get a barcode, as if the y parameter was ignored in the operations ...

BlastingKap · Accepted Answer · 2019-07-31T14:11:35.517

Had to implement this in C recently and this post helped get me started. One fix for the noise function as previously stated.

float noise(int x, int y) {
    int n;

    n = x + y * 57;
    n = (n << 13) ^ n;
    return (1.0 - ( (n * ((n * n * 15731) + 789221) +  1376312589) & 0x7fffffff) / 1073741824.0);
}

Most importantly, is that Fractal Brownian Motion is really important when implementing Perlin Noise for use in a height map. Instead of following the pseudo code by Hugo Elias, I used this Google Code Snippet.

float perlin_two(float x, float y, float gain, int octaves, int hgrid) {
    int i;
    float total = 0.0f;
    float frequency = 1.0f/(float)hgrid;
    float amplitude = gain;
    float lacunarity = 2.0;

    for (i = 0; i < octaves; ++i)
    {
        total += noise_handler((float)x * frequency, (float)y * frequency) * amplitude;         
        frequency *= lacunarity;
        amplitude *= gain;
    } 

    return (total);
}

Before I did this, I'd had the same issue with uniform 'stripping' or just complete random looking height maps.

What's `noise_hander`? Is it just any function which can give you noise? (In which case, what's a `hander`? Why not just call it `noise_generator`?) — Nic, Jul 20 '18 at 04:13

score 0 · Answer 2 · edited May 23 '17 at 11:52

0

You got to debug & rework probably all your variable types; do the math in pen & paper or in a debugger and check the range of the values. e.g. here

noise(int x, int y) {
      int n = x + y * 57;
      n = pow((n << 13), n);  ...
}

n will either underflow or overflow for almost all non-zero values. Good random values can't be produced with any random method, which is what these formulas look like.

Check also this topic for some reference code.

edited May 23 '17 at 11:52

Community

1
1

answered May 15 '13 at 16:45

Aki Suihkonen

19,144
1
36
57

I change the line : "n = pow((n << 13), n);" to "n = (n << 13) ^ n" which was a first error. Now I get bar code. How should I adapt the pseudo code of this function, written on the tutorial, to C language, if it under or overflow ? – tankyx May 15 '13 at 16:57
The bar code is an indication of the selection of the formula `n=x+y*57`; where when you use (x+1) vs (y+1) one of those changes doesn't change the output much (or at all). (x+1) vs. (y+1)*57 ? – Aki Suihkonen May 15 '13 at 18:15
If I change for n = (x*57) + (y * 57), i get some other colors but always bar code, as if I try n = (x + y) * 57. Perlin values are modified, but still draws a cycle ... – tankyx May 15 '13 at 18:31

2D perlin noise in C

2 Answers2