/// <summary>
/// Returns the size x size neighborhood (central pixel included) of the pixel located at (x, y)
/// in the image c. Returns the color in channel (0: Red, 1: Green, 2: Blue).
/// </summary>
/// <param name="c"></param>
/// <param name="x"></param>
/// <param name="y"></param>
/// <param name="size"></param>
/// <param name="channel"></param>
/// <returns></returns>
public static byte[] GetNeighborhood(ColorImage c, int x, int y, int size, int channel)
{
if(channel < 0 || channel > 2)
throw new ArgumentException("Illegal channel");
byte[] ret = new byte[size * size];
int count = 0;
int low = -size / 2;
int high = size / 2;
for(int i = low; i <= high; i++) {
for(int j = low; j <= high; j++) {
switch(channel) {
case 0:
ret[count++] = c.Red[x + i, y + j];
break;
case 1:
ret[count++] = c.Green[x + i, y + j];
break;
case 2:
ret[count++] = c.Blue[x + i, y + j];
break;
}
}
}
return ret;
}
public ColorImage Convolution(ColorImage img, Filter f)
{
int[][] kernel = f.Values;
double scale = f.ScaleFactor;
int offset = f.Offset;
ColorImage ret = new ColorImage(img.Height, img.Width);
if(kernel.Length != kernel[0].Length)
throw new ArgumentException("Kernel must be quadratic in shape (n x n)");
int low = -kernel.Length / 2;
int high = kernel.Length / 2;
for(int i = high; i < img.Height - high; i++) {
for(int j = high; j < img.Width - high; j++) {
int newR = 0, newG = 0, newB = 0;
for(int i_k = low; i_k <= high; i_k++) {
for(int j_k = low; j_k <= high; j_k++) {
// TODO Edge problem
int r_i = 0, g_i = 0, b_i = 0;
r_i = img.Red[i + i_k, j + j_k];
g_i = img.Green[i + i_k, j + j_k];
b_i = img.Blue[i + i_k, j + j_k];
newR += (r_i * kernel[i_k + high][j_k + high]);
newG += (g_i * kernel[i_k + high][j_k + high]);
newB += (b_i * kernel[i_k + high][j_k + high]);
}
}
double r = (double) newR * scale + offset;
double g = (double) newG * scale + offset;
double b = (double) newB * scale + offset;
ret.Red[i, j] = Clamp((int) r, 0, 255);
ret.Green[i, j] = Clamp((int) g, 0, 255);
ret.Blue[i, j] = Clamp((int) b, 0, 255);
}
}
return ret;
}
public MainForm()
{
InitializeComponent();
height = pictureBox1.Height;
width = pictureBox1.Width;
img = new ColorImage(height, width);
bmp = new System.Drawing.Bitmap(width, height, System.Drawing.Imaging.PixelFormat.Format24bppRgb);
comboBox1.SelectedIndex = 0;
selectedFilter = (FilterName) comboBox1.SelectedIndex;
initFilters();
}
public void TestImage()
{
ColorImage img = new ColorImage(15, 15);
for(int i = 0; i < 15; i++) {
for(int j = 0; j < 15; j++) {
img.Red[i, j] = 5;
}
}
var result = ColorImage.GetNeighborhood(img, 1, 1, 3);
var actual = new byte[] { 5, 5, 5, 5, 5, 5, 5, 5, 5};
int k = 0;
foreach(var b in result)
Assert.AreEqual(b, actual[k++]);
k = 0;
foreach(var b in actual)
Assert.AreEqual(b, result[k++]);
}
private void SobelFiter(ColorImage img)
{
var filtered_sobel = Convolution(img, sobel1).ToGrayscaleImage();
ColorImage tmp = new ColorImage(filtered_sobel, filtered_sobel, filtered_sobel);
tmp.SaveToBitmap(filtered_bmp);
this.bmp = filtered_bmp;
this.img = tmp;
pictureBox1.Image = bmp;
}
private void openButton_Click(object sender, EventArgs e)
{
if(openFileDialog1.ShowDialog() == DialogResult.OK) {
bmp = new System.Drawing.Bitmap(openFileDialog1.FileName);
pictureBox1.Image = bmp;
img = new ColorImage(bmp.Height, bmp.Width);
img.LoadFromBitmap(bmp);
}
}
private void applyFilterButton_Click(object sender, EventArgs e)
{
var filtered_bmp = new System.Drawing.Bitmap(img.Width, img.Height, System.Drawing.Imaging.PixelFormat.Format24bppRgb);
switch(selectedFilter) {
case FilterName.Median:
var filtered_median = MedianFilter(img, 3);
filtered_median.SaveToBitmap(filtered_bmp);
this.bmp = filtered_bmp;
this.img = filtered_median;
pictureBox1.Image = bmp;
break;
case FilterName.Average:
var filtered_avg = Convolution(img, average);
filtered_avg.SaveToBitmap(filtered_bmp);
this.bmp = filtered_bmp;
this.img = filtered_avg;
pictureBox1.Image = bmp;
break;
// Gaussian filter
case FilterName.Gauss:
var filtered_gauss = Convolution(img, gauss);
filtered_gauss.SaveToBitmap(filtered_bmp);
this.bmp = filtered_bmp;
this.img = filtered_gauss;
pictureBox1.Image = bmp;
break;
case FilterName.Sobel:
SobelFiter(img);
break;
}
}
public ColorImage MedianFilterFast(ColorImage img, int windowSize)
{
ColorImage ret = new ColorImage(img.Height, img.Width);
int[] hist = new int[256];
// TODO implement Huang's fast median filtering algorithm
return ret;
}
public ColorImage MedianFilter(ColorImage img, int windowSize)
{
ColorImage ret = new ColorImage(img.Height, img.Width);
for(int i = 1; i < img.Height - 1; i++) {
for(int j = 1; j < img.Width - 1; j++) {
var neighborhood_r = ColorImage.GetNeighborhood(img, i, j, windowSize, 0);
var neighborhood_g = ColorImage.GetNeighborhood(img, i, j, windowSize, 1);
var neighborhood_b = ColorImage.GetNeighborhood(img, i, j, windowSize, 2);
Array.Sort(neighborhood_r);
Array.Sort(neighborhood_g);
Array.Sort(neighborhood_b);
ret.Red[i, j] = neighborhood_r[neighborhood_r.Length / 2];
ret.Green[i, j] = neighborhood_g[neighborhood_g.Length / 2];
ret.Blue[i, j] = neighborhood_b[neighborhood_b.Length / 2];
}
}
return ret;
}