🚀 go-pugleaf

In article <7328a61a.0310092311.7ba03fa1@posting.google.com>,
 heroin_c@hotmail.com (wall) wrote:

> I use this code selected the rect which has red eyes.
> 
> // the selected rect which has red eye
> // the selected image part GWorld - theSelectImageGWorld
> LGWorld*	theSelectImageGWorld = NEW LGWorld( theSelectRect,32 );
> ThrowIfNil_( theSelectImageGWorld ); 
> 
> // the orignal image GWorld - theImageGWorld
> PixMapHandle	theOPixMapH = ::GetGWorldPixMap(
> theImageGWorld->GetMacGWorld() );
> PixMapHandle	theSelectPixMapH = ::GetGWorldPixMap(
> theSelectImageGWorld->GetMacGWorld() );
> Ptr theOBaseAddr = ::GetPixBaseAddr( theOPixMapH );
> Ptr theSelectBaseAddr = ::GetPixBaseAddr( theSelectPixMapH );
> short  theORowBytes = ::GetPixRowBytes( theOPixMapH );
> short  theSelectRowBytes = ::GetPixRowBytes( theSelectPixMapH );
> 
> ::LockPixels( theOPixMapH );
> ::LockPixels( theSelectPixMapH );
> {
> 	Ptr		theSrcPtr, theDstPtr;
> 	for(short theY = mSelRectToImageOrignal.top; theY <
> mSelRectToImageOrignal.bottom; theY++)
> 	{
> 		theSrcPtr = theOBaseAddr + theORowBytes * theY;
> 		theSrcPtr += mSelRectToImageOrignal.left << 2;
> 		theDstPtr = theSelectBaseAddr + theSelectRowBytes * ( theY -
> mSelRectToImageOrignal.top );
> 		for (short theX = mSelRectToImageOrignal.left; theX <
> mSelRectToImageOrignal.right; theX++)
> 		{
> 			* ((long*) theDstPtr++) = * ((long*) theSrcPtr++);
> 			* ((long*) theDstPtr++) = * ((long*) theSrcPtr++);
> 			* ((long*) theDstPtr++) = * ((long*) theSrcPtr++);
> 			* ((long*) theDstPtr++) = * ((long*) theSrcPtr++);
> 
> 		}	
> 	}		
> }
> ::UnlockPixels( theOPixMapH );
> ::UnlockPixels( theSelectPixMapH );
> 
> what can I do to removing red eye in the selected image part?

As long as you can be relatively sure that the user has selected a close 
boundary around the pupil, you could just reduce the red level 
everywhere to the average of the green and blue values. This will help 
establish a reasonable intensity for the red channel and will capture 
any residual hilights that might be present. You may still need to bring 
down the overall intensity of each channel by some fixed percentage to 
avoid overly bright pupils.

Each 32-bit pixel is layed out as argb, 8 bits each. You should leave 
the a value alone.

Chris.

In article <7328a61a.0310092311.7ba03fa1@posting.google.com>,
 heroin_c@hotmail.com (wall) wrote:

> I use this code selected the rect which has red eyes.
> 
> // the selected rect which has red eye
> // the selected image part GWorld - theSelectImageGWorld
> LGWorld*	theSelectImageGWorld = NEW LGWorld( theSelectRect,32 );
> ThrowIfNil_( theSelectImageGWorld ); 
> 
> // the orignal image GWorld - theImageGWorld
> PixMapHandle	theOPixMapH = ::GetGWorldPixMap(
> theImageGWorld->GetMacGWorld() );
> PixMapHandle	theSelectPixMapH = ::GetGWorldPixMap(
> theSelectImageGWorld->GetMacGWorld() );
> Ptr theOBaseAddr = ::GetPixBaseAddr( theOPixMapH );
> Ptr theSelectBaseAddr = ::GetPixBaseAddr( theSelectPixMapH );
> short  theORowBytes = ::GetPixRowBytes( theOPixMapH );
> short  theSelectRowBytes = ::GetPixRowBytes( theSelectPixMapH );
> 
> ::LockPixels( theOPixMapH );
> ::LockPixels( theSelectPixMapH );
> {
> 	Ptr		theSrcPtr, theDstPtr;
> 	for(short theY = mSelRectToImageOrignal.top; theY <
> mSelRectToImageOrignal.bottom; theY++)
> 	{
> 		theSrcPtr = theOBaseAddr + theORowBytes * theY;
> 		theSrcPtr += mSelRectToImageOrignal.left << 2;
> 		theDstPtr = theSelectBaseAddr + theSelectRowBytes * ( theY -
> mSelRectToImageOrignal.top );
> 		for (short theX = mSelRectToImageOrignal.left; theX <
> mSelRectToImageOrignal.right; theX++)
> 		{
> 			* ((long*) theDstPtr++) = * ((long*) theSrcPtr++);
> 			* ((long*) theDstPtr++) = * ((long*) theSrcPtr++);
> 			* ((long*) theDstPtr++) = * ((long*) theSrcPtr++);
> 			* ((long*) theDstPtr++) = * ((long*) theSrcPtr++);
> 
> 		}	
> 	}		
> }
> ::UnlockPixels( theOPixMapH );
> ::UnlockPixels( theSelectPixMapH );
> 
> what can I do to removing red eye in the selected image part?

I should add that you may want to leave a pixel alone if its red value 
does not exceed some high threshold. This will prevent discolration of 
the iris's inside the boundary. You could precalculate the threshold by 
histogramming the contents and looking for a sharp increase in the value 
of the red channel. Or you could use a fixed value.

Chris.

Thanks Chris.
But could you give me some code about this.My english is very poor,so
I think maybe I can understand your idea from source code.

In article <7328a61a.0310101735.6ad75c32@posting.google.com>,
 heroin_c@hotmail.com (wall) wrote:

> Thanks Chris.
> But could you give me some code about this.My english is very poor,so
> I think maybe I can understand your idea from source code.

HI,
actually I don't have any code. that was just a general process I was 
describing. It might go something like this:

/* pass 1 -histogram red values */

UInt8 totals[256];
for( each pixel )
   r = (r value)
   totals[r]++

/* analyse histogram -find maximum drop off. this is very crude! */
/* you should definitely find a better way to find the level at */
/* which red eye is occurring. */
UInt32 threshold = 254;
Uint32 delta = 0;
for( i = 0; i < 254; i++ )
   if( delta < totals[i + 1] - totals[i] ) {
      delta = totals[i + 1] - totals[i];
      threshold = i;
      }

/* threshold contains the level at which redeye is deemed to be present*/

/* pass 2 -modify red values */
for( each pixel )
   r = (r value)
   if( r >= threshold )
      (r value) = (b value + g value) / 2

/* pass 2 discards red values over the threshold and replaces them
with the average of b and g */