[bdlt]

take a look at the formula(s) in the 10918-1 spec for DCT & IDCT.

see A.3.3 on page 27

[Advertisements]

I changed the equation but it produced the same values?

public class CosineTransformation {

   public CosineTransformation() {
   }

   public double[][] F = new double[8][8];
   public double Temp = 0.0;
   public double c = 0.0;

   public double[][] DCT(double[][] f) {
	  for(int u = 0; u < 8; u++) {
		 for(int v = 0; v < 8; v++) {
			if(v == 0 && u == 0) {
			   c = 0.5;
			} else {
			   c = 1.0; 
			}

			for(int x = 0; x < 8; x++) {
			   for(int y = 0; y < 8; y++) {
				  Temp += f[x][y] * Math.acos( (((2*x)+1) * u * Math.PI) / 16 )
								  * Math.acos( (((2*y)+1) * v * Math.PI) / 16 );
			   }
			}
			F[u][v] =  (c/4.0) * Temp;
			Temp = 0.0;
		 }
	  }
	  return F;
   }

   public void IDCT(double[][] f) {
   }

}

that follows the equation in the spec(I think)
any Ideas why its not working?

Edited by staticVoid, 28 August 2007 - 07:10 AM.

[staticVoid]

I changed the equation but it produced the same values?

public class CosineTransformation {

   public CosineTransformation() {
   }

   public double[][] F = new double[8][8];
   public double Temp = 0.0;
   public double c = 0.0;

   public double[][] DCT(double[][] f) {
	  for(int u = 0; u < 8; u++) {
		 for(int v = 0; v < 8; v++) {
			if(v == 0 && u == 0) {
			   c = 0.5;
			} else {
			   c = 1.0; 
			}

			for(int x = 0; x < 8; x++) {
			   for(int y = 0; y < 8; y++) {
				  Temp += f[x][y] * Math.acos( (((2*x)+1) * u * Math.PI) / 16 )
								  * Math.acos( (((2*y)+1) * v * Math.PI) / 16 );
			   }
			}
			F[u][v] =  (c/4.0) * Temp;
			Temp = 0.0;
		 }
	  }
	  return F;
   }

   public void IDCT(double[][] f) {
   }

}

that follows the equation in the spec(I think)
any Ideas why its not working?

Edited by staticVoid, 28 August 2007 - 07:10 AM.

[bdlt]

1. use cos() ... not acos()

2. / 16 ... is int division
try ... / 16.0

3. f[x][y]
try ... f[y][x]

4. F[u][v]
try ... F[v][u]

...........................................
not sure about this one

5. u == 0 && v == 0
might be u == 0 || v == 0

Edited by bdlt, 28 August 2007 - 08:00 AM.

[staticVoid]

hey, It worked!
thanx, now just huffman to work out

Edited by staticVoid, 28 August 2007 - 01:30 PM.

[bdlt]

congrats!

the tutorial below decodes a simple 2 block image. all of the AC values are 0.
it covers the basic steps for decoding huffman.

http://www.impulsead...man-coding.html

the most useful part is from the top up to 'conversion to spatial domain'(about half way through the page)

note - the example uses custom huffman tables(not K3-K6 of the spec)

Edited by bdlt, 29 August 2007 - 03:27 PM.

[staticVoid]

is there a method in java for reading bits from a file?

[bdlt]

sorry, I'm not aware of any methods that read bits from a file

but ...

there are bitwise operators and methods

http://www.particle....ter10/bits.html

http://java.sun.com/...dbolts/op3.html

[staticVoid]

I'm not sure if this FDCT is giving the right results

public class CosineTransformation {

   public CosineTransformation() {
   }

   public double[][] F = new double[8][8];
   public double Temp = 0;
   public double c = 0;

   public double[][] DCT(float[][] f) {
	  for(int u = 0; u < 8; u++) {
		 for(int v = 0; v < 8; v++) {
			if(v == 0 && u == 0) {
			   c = 0.5;
			} else {
			   c = 1.0;
			}
			for(int x = 0; x < 8; x++) {
			   for(int y = 0; y < 8; y++) {
				  Temp += f[y][x] * Math.cos( (((2*x)+1) * u * Math.PI) / 16.0 )
								  * Math.cos( (((2*y)+1) * v * Math.PI) / 16.0 );
			   }
			}
			F[v][u] =  (c/4.0) * Temp;
			Temp = 0.0;
		 }
	  }
	  return F;
   }

   public void IDCT(int channel) {
   }

}

could you test it and see if its working?

Heres the the table from http://en.wikipedia.org/wiki/JPEG:

public float[][] TestArrayTwo = 

   {  {-76, -73, -67, -62, -58, -67, -64, -55},
	  {-65, -69, -73, -38, -19, -43, -59, -56},
	  {-66, -69, -60, -15, 16, -24, -62, -55 },
	  {-65, -70, -57, -6, 26, -22, -58, -59  },
	  {-61, -67, -60, -24, -2, -40, -60, -58 },
	  {-49, -63, -68, -58, -51, -60, -70, -53},
	  {-43, -57, -64, -69, -73, -67, -63, -45},
	  {-41, -49, -59, -60, -63, -52, -50, -34}  };

I put the array into the FDCT but the values didnt match the ones in wikipedia.?
I know theres probaly a few errors in the code.

Edited by staticVoid, 03 September 2007 - 10:50 AM.

[bdlt]

evaluation of the code may involve some 'hand' calculations.

please let me know if this becomes resolved before you receive the next reply.

[bdlt]

try this:

replace this
if(v == 0 && u == 0) {
c = 0.5;
} else {
c = 1.0;
}

with:

c = 1.0;
if( u == 0 ) {
c *= Math.sqrt(2.0) / 2.0;
}
if( v == 0 ) {
c *= Math.sqrt(2.0) / 2.0;
}

Edited by bdlt, 03 September 2007 - 12:52 PM.

[staticVoid]

It's still not the right values , I'm getting the equation from http://www.opennet.r...ormats/jpeg.txt (I think it's right)

[bdlt]

look at Temp for u = 0, v = 0

Temp = -3323
c = 0.5

F[0][0] = c * temp / 4.0 = (0.5 * -3323) / 4.0 = -415.375
..............................................................................
other Temp values
(V,U)
0,1 -170.76
0,2 -346.18
0,3 154.08
0,4 317.50
0,5 -113.68
0,6 -13.504
0,7 2.61

1.0 25.26
2,0 -264.93

..........................................................................

please post your current DCT code

Edited by bdlt, 03 September 2007 - 07:09 PM.

[staticVoid]

I'm getting the values you listed in the right order but for example:

F(0, 1) = -170.76 * ( (c=1) /4 ) = -42.675

when (according to wikipedia) it should be -30 ??

Heres my DCT code:

public double[][] DCT(double[][] f) {
	  for(int u = 0; u < 8; u++) {
		 for(int v = 0; v < 8; v++) {
			if(v == 0 && u == 0) {
			   c = 0.5;
			} else {
			   c = 1.0; 
			}
			for(int x = 0; x < 8; x++) {
			   for(int y = 0; y < 8; y++) {
				  Temp += f[y][x] * Math.cos( (((2*x)+1) * u * Math.PI) / 16.0 )
								  * Math.cos( (((2*y)+1) * v * Math.PI) / 16.0 );
			   }
			}
			F[v][u] =  (c/4.0) * Temp;
			Temp = 0.0;
		 }
	  }
	  return F;
   }

I've also realized the values I'm getting are really close to (if not the same as) the table on wikipedia for a lot of the values but some are just way off

Edited by staticVoid, 04 September 2007 - 01:32 AM.

[bdlt]

humor me ... please retry the suggestion in post #55

................................................................

the value for 'c' is a little confusing

avoid using jpeg.txt for DCT

wik and the spec agree on the DCT 'formula'

if we look at the spec(A.3.3 on p27) we see:
Cu, Cv = 1 / sqrt(2) for u,v = 0
otherwise Cu, Cv = 1

the suggestion in post #55 uses sqrt(2) / 2 which is the same as 1/sqrt(2)
sorry ... that probably adds to the confusion

.............................................................

'c' in your code is actually Cu * Cv

we have 4 different sets of Cu and Cv

1. u = 0 and v = 0(the DC value)
c = Cu * Cv = 1/ sqrt(2) * 1/sqrt(2) = 0.5
2. u = 0 and v != 0
c = 1/sqrt(2) * 1 = 1/sqrt(2) = 0.707
3 . u != 0 and v = 0
c = 1 * 1/sqrt(2) = 1/sqrt(2) = 0.707
4. u!= 0 and v!= 0
c = 1 * 1 = 1.0

[staticVoid]

It worked - the reason it never worked last time was because I had display: F[y][x] instead of F[x][y] so it was right all along just not in the right order when I was displaying it.