#include <cglib/rt/bvh.
h>
#include <cglib/rt/triangle_soup.h>
#include <cglib/core/image.h>
#include <complex>
/*
* Create a 1 dimensional normalized gauss kernel
*
* Parameters:
* - sigma: the parameter of the gaussian
* - kernel_size: the size of the kernel (has to be odd)
* - kernel: an array with size kernel_size elements that
* has to be filled with the kernel values
*
*/
void Image::create_gaussian_kernel_1d(
float sigma,
int kernel_size,
float* kernel)
{
cg_assert(kernel_size % 2 == 1);
// TODO: calculate filter values as described on the exercise sheet.
// Make sure your kernel is normalized
int lowestx = -(floor(kernel_size / 2));
float normalize = 0.0;
for (int i = 0; i < kernel_size; ++i) {
kernel[i] = (1.0 / (glm::sqrt(2.0 * glm::pi<float>())* sigma)) *
glm::pow(glm::e<float>(), -((lowestx) * (lowestx) / (2.0* sigma * sigma)));
lowestx++;
}
for (int i = 0; i < kernel_size; ++i) {
normalize += kernel[i];
}
for (int i = 0; i < kernel_size; ++i) {
kernel[i] = kernel[i] / normalize;
}
/*
* Create a 2 dimensional quadratic and normalized gauss kernel
*
* Parameters:
* - sigma: the parameter of the gaussian
* - kernel_size: the size of the kernel (has to be odd)
* - kernel: an array with kernel_size*kernel_size elements that
* has to be filled with the kernel values
*/
void Image::create_gaussian_kernel_2d(
float sigma,
int kernel_size,
float* kernel)
{
cg_assert(kernel_size % 2 == 1);
int lowestnumber = -(floor(kernel_size / 2));
float normalize = 0.0;
for (int j = 0; j < kernel_size; ++j) {
for (int i = 0; i < kernel_size; ++i) {
kernel[i + j * kernel_size] = (1.0 / (2.0 * glm::pi<float>() *
(sigma * sigma)))* glm::pow(glm::e<float>(), -(((lowestnumber + j) * (lowestnumber
+ j) + (lowestnumber + i) * (lowestnumber + i)) / (2.0 * sigma * sigma)));
for (int j = 0; j < kernel_size; ++j) {
for (int i = 0; i < kernel_size; ++i) {
normalize += kernel[i + j*kernel_size];
}
}
for (int j = 0; j < kernel_size; ++j) {
for (int i = 0; i < kernel_size; ++i) {
kernel[i + j * kernel_size] = kernel[i + j*kernel_size] /
normalize;
}
}
}
/*
* Convolve an image with a 2d filter kernel
*
* Parameters:
* - kernel_size: the size of the 2d-kernel
* - kernel: the 2d-kernel with kernel_size*kernel_size elements
* - wrap_mode: needs to be known to handle repeating
* textures correctly
*/
void Image::filter(Image *target, int kernel_size, float* kernel, WrapMode
wrap_mode) const
{
cg_assert(kernel_size % 2 == 1 && "kernel size should be odd.");
cg_assert(kernel_size > 0 && "kernel size should be greater than 0.");
cg_assert(target);
cg_assert(target->getWidth() == m_width && target->getHeight() == m_height);
int lowestnumber = -(floor(kernel_size / 2));
for (int i = 0; i < m_width; ++i) { //rupert hier width statt height
for (int j = 0; j < m_height; ++j) { // height statt width
glm::vec4 newpixel;
for (int a = 0; a < kernel_size; ++a) {
for (int b = 0; b < kernel_size; ++b) {
newpixel += kernel[a + kernel_size*b] * (getPixel(i +
lowestnumber + a, j + lowestnumber + b));
} // a + kernelsize *b nachfragen
}
target->setPixel(i, j, newpixel);
}
}
/*
/*
* Convolve an image with a separable 1d filter kernel
*
* Parameters:
* - kernel_size: the size of the 1d kernel
* - kernel: the 1d-kernel with kernel_size elements
* - wrap_mode: needs to be known to handle repeating
* textures correctly
*/
void Image::filter_separable(Image *target, int kernel_size, float* kernel,
WrapMode wrap_mode) const
{
cg_assert (kernel_size%2==1 && "kernel size should be odd.");
cg_assert (kernel_size > 0 && "kernel size should be greater than 0.");
cg_assert (target);
cg_assert (target->getWidth() == m_width && target->getHeight() == m_height);
Image temp(m_width, m_height);
int lowestnumber = -floor((kernel_size / 2));
for (int i = 0; i < m_width; ++i) {
for (int j = 0; j < m_height; j++) {
glm::vec4 PixelHeight;
for (int k = 0; k < kernel_size; k++) {
PixelHeight += kernel[k] * getPixel(i, j+ lowestnumber + k,
wrap_mode);
} temp.setPixel(i, j, PixelHeight);
}
}
for (int i = 0; i < m_width; ++i) {
for (int j = 0; j < m_height; j++) {
glm::vec4 PixelWidth;
for (int k = 0; k < kernel_size; k++) {
PixelWidth += kernel[k] * temp.getPixel(i+ lowestnumber +
k,j, wrap_mode);
} target->setPixel(i, j, PixelWidth);
}
}
// TODO: realize the 2d convolution with two
// convolutions of the image with a 1d-kernel.
// convolve the image horizontally and then convolve
// the result vertically (or vise-versa).
//
// use the methods getPixel(x, y, wrap_mode) and
// setPixel(x, y, value) to get and set pixels of an image
/**
* Reorder triangle indices in the vector triangle_indices
* in the range [first_triangle_idx, first_triangle_idx+num_triangles-1]
* so that the range is split in two sets where all triangles in the first set
* are "less than equal" than the median, and all triangles in the second set
* are "greater than equal" the median.
*
* Ordering ("less than") is defined by the ordering of triangle
* bounding box centers along the given axis.
*
* Triangle indices within a set need not be sorted.
*
* The resulting sets must have an equal number of elements if num_triangles
* is even. Otherwise, one of the sets must have one more element.
*
* For example, 8 triangles must be split 4-4. 7 Triangles must be split
* 4-3 or 3-4.
*
* Parameters:
* - first_triangle_idx: The index of the first triangle in the given range.
* - num_triangles: The number of triangles in the range.
* - axis: The sort axis. 0 is X, 1 is Y, 2 is Z.
*
* Return value:
* - The number of triangles in the first set.
*/
int BVH::reorder_triangles_median(
int first_triangle_idx,
int num_triangles,
int axis)
{
cg_assert(first_triangle_idx < static_cast<int>(triangle_indices.size()));
cg_assert(first_triangle_idx >= 0);
cg_assert(num_triangles <= static_cast<int>(triangle_indices.size() -
first_triangle_idx));
cg_assert(num_triangles > 1);
cg_assert(axis >= 0);
cg_assert(axis < 3);
// TODO: Implement reordering.
return 0;
}
/*
* Build a BVH recursively using the object median split heuristic.
*
* This method must first fully initialize the current node, and then
* potentially split it.
*
* A node must not be split if it contains MAX_TRIANGLES_IN_LEAF triangles or
* less. No leaf node may be empty. All nodes must have either two or no
* children.
*
* Use reorder_triangles_median to perform the split in triangle_indices.
* Split along X for depth 0. Then, proceed in the order Y, Z, X, Y, Z, X, Y, ..
*
* Parameters:
* - node_idx: The index of the node to be split.
* - first_triangle_idx: An index into the array triangle_indices. It points
* to the first triangle contained in the current node.
* - num_triangles: The number of triangles contained in the current node.
*/
void BVH::
build_bvh(int node_idx, int first_triangle_idx, int num_triangles, int depth)
{
cg_assert(num_triangles > 0);
cg_assert(node_idx >= 0);
cg_assert(node_idx < static_cast<int>(nodes.size()));
cg_assert(depth >= 0);
Node& node = nodes[node_idx];
// TODO: Implement recursive build.
node.triangle_idx = first_triangle_idx;
node.num_triangles = num_triangles;
node.aabb.min = glm::vec3(-FLT_MAX);
node.aabb.max = glm::vec3(FLT_MAX);
node.left = -1;
node.right = -1;
}
/*
* Intersect the BVH recursively, returning the nearest intersection if
* there is one.
*
* Caution: BVH nodes may overlap.
*
* Parameters:
* - ray: The ray to intersect the BVH with.
* - idx: The node to be intersected.
* - nearest_intersection: The distance to the intersection point, if an
* intersection was found. Must not be changed
* otherwise.
* - isect: The intersection, if one was found. Must not be
* changed otherwise.
*
* Return value:
* true if an intersection was found, false otherwise.
*/
bool BVH::
intersect_recursive(const Ray &ray, int idx, float *nearest_intersection,
Intersection* isect) const
{
cg_assert(nearest_intersection);
cg_assert(isect);
cg_assert(idx >= 0);
cg_assert(idx < static_cast<int>(nodes.size()));
const Node &n = nodes[idx];
// This is a leaf node. Intersect all triangles.
if(n.left < 0) {
glm::vec3 bary(0.f);
bool hit = false;
for(int i = 0; i < n.num_triangles; i++) {
int x = triangle_indices[n.triangle_idx + i];
float dist;
glm::vec3 b;
if(intersect_triangle(ray.origin, ray.direction,
triangle_soup.vertices[x * 3 + 0],
triangle_soup.vertices[x * 3 + 1],
triangle_soup.vertices[x * 3 + 2],
b, dist)) {
hit = true;
if(dist <= *nearest_intersection) {
*nearest_intersection = dist;
bary = b;
cg_assert(x >= 0);
if(isect)
triangle_soup.fill_intersection(isect, x,
*nearest_intersection, bary);
}
}
}
return hit;
}
// This is an inner node. Recurse into child nodes.
else {
// TODO: Implement recursive traversal here.
}
return false;
}
// Spectrum and reconstruction are 2D memory blocks of the same size [sx, sy]
// in row major layout, i.e one row of size sx after the other.
//
// Spectrum contains the complex fourier-coefficients \hat{x}_{kl}.
//
// Reconstruct the original grayscale image using fourier transformation!
void DiscreteFourier2D::reconstruct(
int sx, int sy,
std::complex<float> const* spectrum,
std::complex<float> * reconstruction)
{
}
// CG_REVISION 3d6fdd132b9492fa6a480cc37c1514dd9997eedf
