// Test the methods
/*
* z = [139 144 149 153 155 155 155 155;
* 144 151 153 156 159 156 156 156;
* 150 155 160 163 158 156 156 156;
* 159 161 162 160 160 159 159 159;
* 159 160 161 162 162 155 155 155;
* 161 161 161 161 160 157 157 157;
* 162 162 161 163 162 157 157 157;
* 162 162 161 161 163 158 158 158];
*
* octave:55> x = dct(z)
* x =
* 436.99 444.06 448.31 452.19 452.19 443 443 443
* -21.818 -14.969 -9.3908 -6.4728 -5.921 -1.7745 -1.7745 -1.7745
* -8.8097 -7.5031 -7.3446 -4.6522 -1.2737 -0.46194 -0.46194 -0.46194
* -2.4118 -3.7457 -3.9958 -3.0683 -1.4969 -1.7704 -1.7704 -1.7704
* 0.70711 -0.70711 -0.70711 -2.4749 0.35355 0.35355 0.35355 0.35355
* 1.365 0.22465 0.90791 0.57409 -1.7777 1.2224 1.2224 1.2224
* -0.94311 -1.4843 0.74614 0.77897 -2.1512 -0.19134 -0.19134 -0.19134
* -1.8165 -1.685 0.14561 2.9764 0.20231 -2.3922 -2.3922 -2.3922
*
* octave:56> idct(x)
* ans =
* 139 144 149 153 155 155 155 155
* 144 151 153 156 159 156 156 156
* 150 155 160 163 158 156 156 156
* 159 161 162 160 160 159 159 159
* 159 160 161 162 162 155 155 155
* 161 161 161 161 160 157 157 157
* 162 162 161 163 162 157 157 157
* 162 162 161 161 163 158 158 158
*/
public static void test()
{
double[][] vals = new double[][] {
new double[] { 139.0, 144.0, 149.0, 153.0, 155.0, 155.0, 155.0, 155.0 },
new double[] { 144.0, 151.0, 153.0, 156.0, 159.0, 156.0, 156.0, 156.0 },
new double[] { 150.0, 155.0, 160.0, 163.0, 158.0, 156.0, 156.0, 156.0 },
new double[] { 159.0, 161.0, 162.0, 160.0, 160.0, 159.0, 159.0, 159.0 },
new double[] { 159.0, 160.0, 161.0, 162.0, 162.0, 155.0, 155.0, 155.0 },
new double[] { 161.0, 161.0, 161.0, 161.0, 160.0, 157.0, 157.0, 157.0 },
new double[] { 162.0, 162.0, 161.0, 163.0, 162.0, 157.0, 157.0, 157.0 },
new double[] { 162.0, 162.0, 161.0, 161.0, 163.0, 158.0, 158.0, 158.0 }
};
DctMethods.PrintMatrix(vals);
DctComirva dctCom = new DctComirva(vals.Length, vals[0].Length);
// dct
double[][] dctVals = dctCom.dct(vals);
DctMethods.PrintMatrix(dctVals);
// idct
double[][] idctVals = dctCom.idct(dctVals);
DctMethods.PrintMatrix(idctVals);
}
// test 1
// Octave results:
// format short g
// z = [1 2 3; 4 5 6; 7 8 9; 10 11 12];
//
//octave:5> dct(z)
//ans =
// 11 13 15
// -6.6913 -6.6913 -6.6913
// 0 0 0
// -0.47554 -0.47554 -0.47554
//
//octave:6> dct2(z)
//ans =
// 22.517 -2.8284 0
// -11.59 0 0
// 0 0 0
// -0.82366 0 0
//
#endregion
private static void Test1(bool _2D = true)
{
var vals = new double[][] {
new double[] { 1.0, 2.0, 3.0 },
new double[] { 4.0, 5.0, 6.0 },
new double[] { 7.0, 8.0, 9.0 },
new double[] { 10.0, 11.0, 12.0 }
};
const double offset = 0.0;
Console.WriteLine("vals: ");
DctMethods.PrintMatrix(vals);
Stopwatch stopWatch = Stopwatch.StartNew();
long startS = stopWatch.ElapsedTicks;
double[][] result;
if (_2D)
{
result = DctMethods.dct2(vals, offset);
Console.WriteLine("dct2 result: ");
}
else
{
result = DctMethods.dct(vals, offset);
Console.WriteLine("dct result: ");
}
long endS = stopWatch.ElapsedTicks;
DctMethods.PrintMatrix(result);
Console.WriteLine("time in ticks: " + (endS - startS));
//result = Filter(result, 1.0);
//Console.WriteLine("dct2 filtered: ");
//PrintMatrix(result);
long startE = stopWatch.ElapsedTicks;
double[][] ivals;
if (_2D)
{
ivals = DctMethods.idct2(result, -offset);
Console.WriteLine("idct2 result: ");
}
else
{
ivals = DctMethods.idct(result, -offset);
Console.WriteLine("idct result: ");
}
long endE = stopWatch.ElapsedTicks;
DctMethods.PrintMatrix(ivals);
Console.WriteLine("Time in ticks: " + (endE - startE));
Assert.Equal(vals, ivals, new JaggedDoubleComparer(0.001));
// Assert.That(ivals, Is.EqualTo(vals).AsCollection.Within(0.001), "fail at [0]");
}
// test 2
// Octave results:
// format short g
/*
* octave:43>
* z = [139 144 149 153 155 155 155 155;
* 144 151 153 156 159 156 156 156;
* 150 155 160 163 158 156 156 156;
* 159 161 162 160 160 159 159 159;
* 159 160 161 162 162 155 155 155;
* 161 161 161 161 160 157 157 157;
* 162 162 161 163 162 157 157 157;
* 162 162 161 161 163 158 158 158];
*
* octave:43> g = dct2(z-128)
* g =
* 235.62 -1.0333 -12.081 -5.2029 2.125 -1.6724 -2.708 1.3238
* -22.59 -17.484 -6.2405 -3.1574 -2.8557 -0.069456 0.43417 -1.1856
* -10.949 -9.2624 -1.5758 1.5301 0.20295 -0.94186 -0.56694 -0.062924
* -7.0816 -1.9072 0.22479 1.4539 0.89625 -0.079874 -0.042291 0.33154
* -0.625 -0.83811 1.4699 1.5563 -0.125 -0.66099 0.60885 1.2752
* 1.7541 -0.20286 1.6205 -0.34244 -0.77554 1.4759 1.041 -0.99296
* -1.2825 -0.35995 -0.31694 -1.4601 -0.48996 1.7348 1.0758 -0.76135
* -2.5999 1.5519 -3.7628 -1.8448 1.8716 1.2139 -0.56788 -0.44564
*
* octave:44> idct2(g)+128
* ans =
* 139 144 149 153 155 155 155 155
* 144 151 153 156 159 156 156 156
* 150 155 160 163 158 156 156 156
* 159 161 162 160 160 159 159 159
* 159 160 161 162 162 155 155 155
* 161 161 161 161 160 157 157 157
* 162 162 161 163 162 157 157 157
* 162 162 161 161 163 158 158 158
*
* octave:49> g = dct(z-128)
* g =
*
* 74.953 82.024 86.267 90.156 90.156 80.964 80.964 80.964
* -21.818 -14.969 -9.3908 -6.4728 -5.921 -1.7745 -1.7745 -1.7745
* -8.8097 -7.5031 -7.3446 -4.6522 -1.2737 -0.46194 -0.46194 -0.46194
* -2.4118 -3.7457 -3.9958 -3.0683 -1.4969 -1.7704 -1.7704 -1.7704
* 0.70711 -0.70711 -0.70711 -2.4749 0.35355 0.35355 0.35355 0.35355
* 1.365 0.22465 0.90791 0.57409 -1.7777 1.2224 1.2224 1.2224
* -0.94311 -1.4843 0.74614 0.77897 -2.1512 -0.19134 -0.19134 -0.19134
* -1.8165 -1.685 0.14561 2.9764 0.20231 -2.3922 -2.3922 -2.3922
*
* octave:50> idct(g)+128
* ans =
*
* 139 144 149 153 155 155 155 155
* 144 151 153 156 159 156 156 156
* 150 155 160 163 158 156 156 156
* 159 161 162 160 160 159 159 159
* 159 160 161 162 162 155 155 155
* 161 161 161 161 160 157 157 157
* 162 162 161 163 162 157 157 157
* 162 162 161 161 163 158 158 158
*/
#endregion
public static void Test2(bool _2D = true, bool random = false)
{
double[][] vals;
if (random)
{
// Generate random integers between 0 and 255
const int N = 8;
var generator = new Random();
vals = new double[N][];
int val;
for (int x = 0; x < N; x++)
{
vals[x] = new double[N];
for (int y = 0; y < N; y++)
{
val = generator.Next(255);
vals[x][y] = val;
}
}
}
else
{
vals = new double[][] {
new double[] { 139.0, 144.0, 149.0, 153.0, 155.0, 155.0, 155.0, 155.0 },
new double[] { 144.0, 151.0, 153.0, 156.0, 159.0, 156.0, 156.0, 156.0 },
new double[] { 150.0, 155.0, 160.0, 163.0, 158.0, 156.0, 156.0, 156.0 },
new double[] { 159.0, 161.0, 162.0, 160.0, 160.0, 159.0, 159.0, 159.0 },
new double[] { 159.0, 160.0, 161.0, 162.0, 162.0, 155.0, 155.0, 155.0 },
new double[] { 161.0, 161.0, 161.0, 161.0, 160.0, 157.0, 157.0, 157.0 },
new double[] { 162.0, 162.0, 161.0, 163.0, 162.0, 157.0, 157.0, 157.0 },
new double[] { 162.0, 162.0, 161.0, 161.0, 163.0, 158.0, 158.0, 158.0 }
};
}
const double offset = -128.0;
Console.WriteLine("vals: ");
DctMethods.PrintMatrix(vals);
Stopwatch stopWatch = Stopwatch.StartNew();
long startS = stopWatch.ElapsedTicks;
double[][] result;
if (_2D)
{
result = DctMethods.dct2(vals, offset);
Console.WriteLine("dct2 result: ");
}
else
{
result = DctMethods.dct(vals, offset);
Console.WriteLine("dct result: ");
}
long endS = stopWatch.ElapsedTicks;
DctMethods.PrintMatrix(result);
Console.WriteLine("Time in ticks: " + (endS - startS));
//result = Filter(result, 0.25);
//result = CutLeastSignificantCoefficients(result);
//Console.WriteLine("dct2 filtered: ");
//PrintMatrix(result);
long startE = stopWatch.ElapsedTicks;
double[][] ivals;
if (_2D)
{
ivals = DctMethods.idct2(result, -offset);
Console.WriteLine("idct2 result: ");
}
else
{
ivals = DctMethods.idct(result, -offset);
Console.WriteLine("idct result: ");
}
long endE = stopWatch.ElapsedTicks;
DctMethods.PrintMatrix(ivals);
Console.WriteLine("Time in ticks: " + (endE - startE));
Assert.That(ivals, Is.EqualTo(vals).AsCollection.Within(0.001), "fail at [0]");
}
public void TestDctMatrix()
{
#region Output from Octave
/*
* z = [139 144 149 153 155 155 155 155;
* 144 151 153 156 159 156 156 156;
* 150 155 160 163 158 156 156 156;
* 159 161 162 160 160 159 159 159;
* 159 160 161 162 162 155 155 155;
* 161 161 161 161 160 157 157 157;
* 162 162 161 163 162 157 157 157;
* 162 162 161 161 163 158 158 158];
*
* octave:55> x = dct(z)
* x =
* 436.99 444.06 448.31 452.19 452.19 443 443 443
* -21.818 -14.969 -9.3908 -6.4728 -5.921 -1.7745 -1.7745 -1.7745
* -8.8097 -7.5031 -7.3446 -4.6522 -1.2737 -0.46194 -0.46194 -0.46194
* -2.4118 -3.7457 -3.9958 -3.0683 -1.4969 -1.7704 -1.7704 -1.7704
* 0.70711 -0.70711 -0.70711 -2.4749 0.35355 0.35355 0.35355 0.35355
* 1.365 0.22465 0.90791 0.57409 -1.7777 1.2224 1.2224 1.2224
* -0.94311 -1.4843 0.74614 0.77897 -2.1512 -0.19134 -0.19134 -0.19134
* -1.8165 -1.685 0.14561 2.9764 0.20231 -2.3922 -2.3922 -2.3922
*
* octave:56> idct(x)
* ans =
* 139 144 149 153 155 155 155 155
* 144 151 153 156 159 156 156 156
* 150 155 160 163 158 156 156 156
* 159 161 162 160 160 159 159 159
* 159 160 161 162 162 155 155 155
* 161 161 161 161 160 157 157 157
* 162 162 161 163 162 157 157 157
* 162 162 161 161 163 158 158 158
*/
#endregion
var vals = new double[][] {
new double[] { 139.0, 144.0, 149.0, 153.0, 155.0, 155.0, 155.0, 155.0 },
new double[] { 144.0, 151.0, 153.0, 156.0, 159.0, 156.0, 156.0, 156.0 },
new double[] { 150.0, 155.0, 160.0, 163.0, 158.0, 156.0, 156.0, 156.0 },
new double[] { 159.0, 161.0, 162.0, 160.0, 160.0, 159.0, 159.0, 159.0 },
new double[] { 159.0, 160.0, 161.0, 162.0, 162.0, 155.0, 155.0, 155.0 },
new double[] { 161.0, 161.0, 161.0, 161.0, 160.0, 157.0, 157.0, 157.0 },
new double[] { 162.0, 162.0, 161.0, 163.0, 162.0, 157.0, 157.0, 157.0 },
new double[] { 162.0, 162.0, 161.0, 161.0, 163.0, 158.0, 158.0, 158.0 }
};
DctMethods.PrintMatrix(vals);
var dctCom = new DctMatrix(vals.Length, vals[0].Length);
// dct
double[][] dctVals = dctCom.Dct(vals);
DctMethods.PrintMatrix(dctVals);
// idct
double[][] idctVals = dctCom.InverseDct(dctVals);
DctMethods.PrintMatrix(idctVals);
Assert.That(idctVals, Is.EqualTo(vals).AsCollection.Within(0.001), "fail at [0]");
}
/// <summary>
/// Calcutate the perceptual hash of an image according to the algorithm given by Dr. Neal Krawetz
/// on his blog: http://www.hackerfactor.com/blog/index.php?/archives/432-Looks-Like-It.html.
/// </summary>
/// <param name="image">The image to hash.</param>
/// <returns>Returns a 'binary string' (aka bitstring) (like. 001010111011100010) which is easy to do a hamming distance on.</returns>
private string GetHash(Bitmap img)
{
// 1. Reduce size.
// Like Average Hash, pHash starts with a small image.
// However, the image is larger than 8x8; 32x32 is a good size.
// This is really done to simplify the DCT computation and not
// because it is needed to reduce the high frequencies.
var simg = CommonUtils.ImageUtils.Resize(img, size, size);
// 2. Reduce color.
// The image is reduced to a grayscale just to further simplify
// the number of computations.
var gimg = CommonUtils.ImageUtils.MakeGrayscale(simg);
double[][] vals = new double[size][];
// for faster pixels access
// http://csharpexamples.com/fast-image-processing-c/
unsafe
{
BitmapData bitmapData = gimg.LockBits(new Rectangle(0, 0, gimg.Width, gimg.Height), ImageLockMode.ReadWrite, gimg.PixelFormat);
int bytesPerPixel = System.Drawing.Bitmap.GetPixelFormatSize(gimg.PixelFormat) / 8;
int heightInPixels = bitmapData.Height;
int widthInBytes = bitmapData.Width * bytesPerPixel;
byte *PtrFirstPixel = (byte *)bitmapData.Scan0;
for (int x = 0; x < widthInBytes; x = x + bytesPerPixel)
{
vals[x / 4] = new double[size];
for (int y = 0; y < heightInPixels; y++)
{
byte *currentLine = PtrFirstPixel + (y * bitmapData.Stride);
// Console.WriteLine("x ({0}) y ({1})", x, y);
// when the image is grayscale RGB has the same value
vals[x / 4][y] = currentLine[x];
}
}
gimg.UnlockBits(bitmapData);
simg.Dispose();
gimg.Dispose();
}
// 3. Calcutate the DCT.
// The DCT separates the image into a collection of frequencies
// and scalars. While JPEG uses an 8x8 DCT, this algorithm uses
// a 32x32 DCT.
double[][] dctVals = DctMethods.dct2(vals);
//double[][] dctVals = DctMethods.idct2(vals);
// 4. Reduce the DCT.
// This is the magic step. While the DCT is 32x32, just keep the
// top-left 8x8. Those represent the lowest frequencies in the
// picture.
// 5 a) Calcutate the average value.
// Like the Average Hash, compute the mean DCT value (using only
// the 8x8 DCT low-frequency values and excluding the first term
// since the DC coefficient can be significantly different from
// the other values and will throw off the average).
double total = 0;
for (int x = 0; x < smallerSize; x++)
{
for (int y = 0; y < smallerSize; y++)
{
total += dctVals[x][y];
}
}
total -= dctVals[0][0];
// 5. b) Calcutate the average value.
double avg = total / (double)((smallerSize * smallerSize) - 1);
// 6. Further reduce the DCT.
// This is the magic step. Set the 64 hash bits to 0 or 1
// depending on whether each of the 64 DCT values is above or
// below the average value. The result doesn't tell us the
// actual low frequencies; it just tells us the very-rough
// relative scale of the frequencies to the mean. The result
// will not vary as long as the overall structure of the image
// remains the same; this can survive gamma and color histogram
// adjustments without a problem.
string hash = String.Empty;
for (int x = 0; x < smallerSize; x++)
{
for (int y = 0; y < smallerSize; y++)
{
if (x != 0 && y != 0)
{
hash += (dctVals[x][y] > avg ? "1" : "0");
}
}
}
return(hash);
}
/// <summary>
/// Computes the perceptual hash of an image according to the algorithm given by Dr. Neal Krawetz
/// on his blog: http://www.hackerfactor.com/blog/index.php?/archives/432-Looks-Like-It.html.
/// </summary>
/// <param name="img">The image to hash.</param>
/// <returns>Returns a 'binary string' (aka bitstring) (like. 001010111011100010) which is easy to do a hamming distance on.</returns>
public string GetHash(Bitmap img)
{
#if DEBUG
img.Save("ImagePHash 1-orig.png");
#endif
// 1. Reduce size.
// Like Average Hash, pHash starts with a small image.
// However, the image is larger than 8x8; 32x32 is a good size.
// This is really done to simplify the DCT computation and not
// because it is needed to reduce the high frequencies.
img = ImageUtils.Resize(img, size, size);
#if DEBUG
img.Save("ImagePHash 2-reduced.png");
#endif
// 2. Reduce color.
// The image is reduced to a grayscale just to further simplify
// the number of computations.
img = ImageUtils.MakeGrayscale(img);
#if DEBUG
img.Save("ImagePHash 3-grayscale.png");
#endif
var vals = new double[size][];
for (int x = 0; x < img.Width; x++)
{
vals[x] = new double[size];
for (int y = 0; y < img.Height; y++)
{
// when the image is grayscale RGB has the same value
vals[x][y] = img.GetPixel(x, y).B;
}
}
#if DEBUG
// create byte array to be able to save the image
var grayscaleByteArray = new byte[size * size];
for (int x = 0; x < size; x++)
{
for (int y = 0; y < size; y++)
{
// add to byte array
grayscaleByteArray[x + (y * size)] = Convert.ToByte(vals[x][y]);
}
}
Image grayscale = ImageUtils.ByteArrayGrayscaleToImage(grayscaleByteArray, size, size);
grayscale.Save("ImagePHash 4-grayscale-array.png");
#endif
// 3. Compute the DCT.
// The DCT separates the image into a collection of frequencies
// and scalars. While JPEG uses an 8x8 DCT, this algorithm uses
// a 32x32 DCT.
double[][] dctVals = DctMethods.dct2(vals);
#if DEBUG
// create image array to be able to save DCT image
// array to keep only the highest frequency items
var dctValsOnlyHighFreq = new double[size][];
// Compressing Range By taking Log
var dctLogVals = new double[dctVals.Length][];
for (int x = 0; x < size; x++)
{
dctValsOnlyHighFreq[x] = new double[dctVals[x].Length];
dctLogVals[x] = new double[dctVals[x].Length];
for (int y = 0; y < size; y++)
{
dctLogVals[x][y] = Math.Log(1 + Math.Abs((int)dctVals[x][y]));
}
}
// Normalizing Array
double dctMin = 0;
double dctMax = 0;
MathUtils.ComputeMinAndMax(dctLogVals, out dctMin, out dctMax);
var dctPixels = new byte[size * size];
for (int x = 0; x < size; x++)
{
for (int y = 0; y < size; y++)
{
// Constrain Range between 0 and 255
int dctPixelVal = (int)(((float)(dctLogVals[x][y] - dctMin) / (float)(dctMax - dctMin)) * 255);
dctPixels[x + (y * size)] = Convert.ToByte(dctPixelVal);
}
}
Image dctImage = ImageUtils.ByteArrayGrayscaleToImage(dctPixels, size, size);
dctImage.Save("ImagePHash 5-dct.png");
#endif
// 4. Reduce the DCT.
// This is the magic step. While the DCT is 32x32, just keep the
// top-left 8x8. Those represent the lowest frequencies in the
// picture.
#if DEBUG
// create image array to be able to save DCT image
var dctPixelsSmallerSize = new byte[smallerSize * smallerSize];
#endif
// 5 a) Compute the average value.
// Like the Average Hash, compute the mean DCT value (using only
// the 8x8 DCT low-frequency values and excluding the first term
// since the DC coefficient can be significantly different from
// the other values and will throw off the average).
double total = 0;
for (int x = 0; x < smallerSize; x++)
{
for (int y = 0; y < smallerSize; y++)
{
#if DEBUG
dctValsOnlyHighFreq[x][y] = dctVals[x][y]; // store in new dct val array
// convert to pixel values
int dctPixelSmallerVal = (int)(((float)(dctLogVals[x][y] - dctMin) / (float)(dctMax - dctMin)) * 255);
dctPixelsSmallerSize[x + (y * smallerSize)] = Convert.ToByte(dctPixelSmallerVal);
#endif
total += dctVals[x][y];
}
}
total -= dctVals[0][0];
#if DEBUG
Image dctImageSmaller = ImageUtils.ByteArrayGrayscaleToImage(dctPixelsSmallerSize, smallerSize, smallerSize);
dctImageSmaller.Save("ImagePHash 6-dct-smaller.png");
// Inverse DCT
double[][] inverseDctVals = DctMethods.idct2(dctValsOnlyHighFreq);
var idctPixels = new byte[size * size];
for (int x = 0; x < size; x++)
{
for (int y = 0; y < size; y++)
{
double inverseDctVal = inverseDctVals[x][y];
inverseDctVal = inverseDctVal < 0 ? 0 : inverseDctVal;
inverseDctVal = inverseDctVal > 255 ? 255 : inverseDctVal;
idctPixels[x + (y * size)] = Convert.ToByte(inverseDctVal);
}
}
Image idctImage = ImageUtils.ByteArrayGrayscaleToImage(idctPixels, size, size);
idctImage.Save("ImagePHash 7-idct.png");
#endif
// 5. b) Compute the average value.
double avg = total / (double)((smallerSize * smallerSize) - 1);
// 6. Further reduce the DCT.
// This is the magic step. Set the 64 hash bits to 0 or 1
// depending on whether each of the 64 DCT values is above or
// below the average value. The result doesn't tell us the
// actual low frequencies; it just tells us the very-rough
// relative scale of the frequencies to the mean. The result
// will not vary as long as the overall structure of the image
// remains the same; this can survive gamma and color histogram
// adjustments without a problem.
string hash = "";
for (int x = 0; x < smallerSize; x++)
{
for (int y = 0; y < smallerSize; y++)
{
if (x != 0 && y != 0)
{
hash += (dctVals[x][y] > avg ? "1" : "0");
}
}
}
return(hash);
}
