| View previous topic :: View next topic |
| Author |
Message |
eljainc
Guest
|
 Posted: Fri Oct 17, 2008 8:24 pm Post subject: Rotation of color bitmap |
 |
|
Hello,
I am running into a bit of difficulty. I have an algorithm that
rotates an 8-bit bitmap (matrix). The content has been grayscale
images in the past. When I apply this towards an 8-bit color image,
the results are less than ideal. It seems the pixels that result look
almost random.
Here is my procedure:
Why would the color objects not be translated properly?
Thanks
Mike
Procedure RotateObject(M1:MatrixPtr;Angle:Single);
Type Small_Image=Array[0..119,0..119] of byte;
Small_ImagePtr=^Small_Image;
Var
i,j,n,m,ci,cj:Integer;
sina,cosa,oldi,oldj:Single;
r_oldi,r_oldj:real;
v,t,xmax,xmin,alpha,beta:Single;
ii,jj:Integer;
{X,rim,newimage:Small_ImagePtr;}
X,rim:Small_ImagePtr;
Const
MM:integer= 120;
NN:integer= 120;
Function Bilinear( x:Small_ImagePtr; a,b:Single; ii,jj:integer ):
Integer;
Var y:Single;
Begin
if ((ii<1) or (ii > 118)) then bilinear:=64 else
if ((jj<1) or (jj> 118)) then bilinear:=64 else
if (x^[ii,jj]=64) and (x^[ii+1,jj+1]=64) and (x^[ii+1,jj]=64) and
(x^[ii,jj+1]=64)
then bilinear:=64 else
{if ((ii<1) or (ii > 118)) then bilinear:=64 else
if ((jj<1) or (jj> 118)) then bilinear:=64 else
}
if ((a=0) and (b=0)) then bilinear:=x^[ii,jj] else
if (a=0) then
begin
y:=int((1-b)*x^[ii,jj]+b*x^[ii,jj+1]+0.5);
if y>63 then bilinear:=63 else bilinear:=round(y);
end else
if (b=0) then
begin
y:=int((1-a)*x^[ii,jj]+a*x^[ii+1,jj]+0.5);
if y>63 then bilinear:=63 else bilinear:=round(y);
end else
begin
y:=(1-a)*(1-b)*x^[ii,jj]+(1-a)*b*x^[ii,jj+1];
y:=y+a*(1-b)*x^[ii+1,jj]+a*b*x^[ii+1,jj+1]+0.5;
if y>63 then y:=63;
bilinear:=round(int(y));
end;
End;
Begin
angle:=6.283185307-angle;
sina:=sin(angle);
cosa:=cos(angle);
ci:=NN div 2;
cj:=MM div 2;
New(x);
For i:=0 to 119 do for j:=0 to 119 do X^[i,j]:=M1^[i-60,j-60];
New(rim);
For i:=0 to (NN-1) do For j:=0 to (MM-1) do
Begin
oldi:=(i-ci)*cosa-(j-cj)*sina+ci;
oldj:=(i-ci)*sina+(j-cj)*cosa+cj;
ii:=round(int(oldi));
jj:=round(int(oldj));
alpha:=oldi-int(oldi);
beta:=oldj-int(oldj);
rim^[i,j]:=bilinear(x,alpha,beta,ii,jj);
end;
for i:=0 to (NN-1) do for j:=0 to (MM-1) do
M1^[i-60,j-60]:=rim^[i,j];
dispose(x);
dispose(rim);
end; |
|
| |
|
 Back to top |
ImageAnalyst
Guest
|
| Posted: Fri Oct 17, 2008 8:45 pm Post subject: Re: Rotation of color bitmap |
|
|
On Oct 17, 4:24 pm, eljainc <elja...@sbcglobal.net> wrote:
[quote]Hello,
I am running into a bit of difficulty. I have an algorithm that
rotates an 8-bit bitmap (matrix). The content has been grayscale
images in the past. When I apply this towards an 8-bit color image,
the results are less than ideal. It seems the pixels that result look
almost random.
Here is my procedure:
Why would the color objects not be translated properly?
Thanks
Mike
Procedure RotateObject(M1:MatrixPtr;Angle:Single);
Type Small_Image=Array[0..119,0..119] of byte;
Small_ImagePtr=^Small_Image;
Var
i,j,n,m,ci,cj:Integer;
sina,cosa,oldi,oldj:Single;
r_oldi,r_oldj:real;
v,t,xmax,xmin,alpha,beta:Single;
ii,jj:Integer;
{X,rim,newimage:Small_ImagePtr;}
X,rim:Small_ImagePtr;
Const
MM:integer= 120;
NN:integer= 120;
Function Bilinear( x:Small_ImagePtr; a,b:Single; ii,jj:integer ):
Integer;
Var y:Single;
Begin
if ((ii<1) or (ii > 118)) then bilinear:=64 else
if ((jj<1) or (jj> 118)) then bilinear:=64 else
if (x^[ii,jj]=64) and (x^[ii+1,jj+1]=64) and (x^[ii+1,jj]=64) and
(x^[ii,jj+1]=64)
then bilinear:=64 else
{if ((ii<1) or (ii > 118)) then bilinear:=64 else
if ((jj<1) or (jj> 118)) then bilinear:=64 else
}
if ((a=0) and (b=0)) then bilinear:=x^[ii,jj] else
if (a=0) then
begin
y:=int((1-b)*x^[ii,jj]+b*x^[ii,jj+1]+0.5);
if y>63 then bilinear:=63 else bilinear:=round(y);
end else
if (b=0) then
begin
y:=int((1-a)*x^[ii,jj]+a*x^[ii+1,jj]+0.5);
if y>63 then bilinear:=63 else bilinear:=round(y);
end else
begin
y:=(1-a)*(1-b)*x^[ii,jj]+(1-a)*b*x^[ii,jj+1];
y:=y+a*(1-b)*x^[ii+1,jj]+a*b*x^[ii+1,jj+1]+0.5;
if y>63 then y:=63;
bilinear:=round(int(y));
end;
End;
Begin
angle:=6.283185307-angle;
sina:=sin(angle);
cosa:=cos(angle);
ci:=NN div 2;
cj:=MM div 2;
New(x);
For i:=0 to 119 do for j:=0 to 119 do X^[i,j]:=M1^[i-60,j-60];
New(rim);
For i:=0 to (NN-1) do For j:=0 to (MM-1) do
Begin
oldi:=(i-ci)*cosa-(j-cj)*sina+ci;
oldj:=(i-ci)*sina+(j-cj)*cosa+cj;
ii:=round(int(oldi));
jj:=round(int(oldj));
alpha:=oldi-int(oldi);
beta:=oldj-int(oldj);
rim^[i,j]:=bilinear(x,alpha,beta,ii,jj);
end;
for i:=0 to (NN-1) do for j:=0 to (MM-1) do
M1^[i-60,j-60]:=rim^[i,j];
dispose(x);
dispose(rim);
end;
[/quote]
I understand images a lot faster than code. Could you post the input
and output gray scale and color images somewhere? |
|
| |
|
| Back to top |
Martin Leese
Guest
|
| Posted: Sat Oct 18, 2008 2:48 am Post subject: Re: Rotation of color bitmap |
|
|
eljainc wrote:
[quote]Hello,
I am running into a bit of difficulty. I have an algorithm that
rotates an 8-bit bitmap (matrix). The content has been grayscale
images in the past. When I apply this towards an 8-bit color image,
the results are less than ideal. It seems the pixels that result look
almost random.
Here is my procedure:
Why would the color objects not be translated properly?
....
Function Bilinear( x:Small_ImagePtr; a,b:Single; ii,jj:integer ):
[/quote]
Er ... because you are using bi-linear
interpolation. Imagine you have neighbouring
pixels with values of 32 and 34. When
interpolated during your rotation, your new
image could well contain a pixel with a value
of 33.
With shades of gray, a graytone of 33 looks
very similar to graytones of 32 and 34.
However, when passed through a palette, 33
could appear very different from 32 or 34.
(I assume you are using a palette, although
using 3, 3, 2 bits for red, green, blue will
behave similarly.)
If I have diagnosed the problem correctly
then the simple solution is to use nearest
neighbour interpolation.
--
Regards,
Martin Leese
E-mail: please@see.Web.for.e-mail.INVALID
Web: http://members.tripod.com/martin_leese/ |
|
| |
|
| Back to top |
|
