Resizing transparent bitmaps with the BlackBerry JDE
Here’s a post I made yesterday on the BlackBerry’s Java Development forum. See the original post here. I updated the code since then, the latest version can be found below.
I recently tried to use Bitmap.scaleInto() to resize transparent bitmaps in my application. Although this function is a nice addition to JDE 5, it tends to produce bad-looking results on transparent bitmaps such as PNG24 files. Forum posts I found on the web for this problem talk about using Bitmap.createAlpha() which does restore transparency but still produces artifacts in the final bitmap such as white edges and wicked pixel colors.
To solve this without the use of an external library, I created the following function. It still uses scaleInto() but in a different manner. I added some comments in the function and created a screenshot showing before and after results (don’t mind the pictures, they’re part of the project I’m currently working on). I’m posting the result here in hope it will help other people searching to achieve the same thing. The code was tested in JDE 5 and JDE 6 and the sample pictures were created using different combinations of filter types and aspect ratios.
package com.patchou.ui;
import net.rim.device.api.system.Bitmap;
import net.rim.device.api.ui.Color;
import net.rim.device.api.ui.Graphics;
/**
*
* @author Patchou
* @version 1.01
*
*/
public class GPATools
{
/**
* Resizes a bitmap with an alpha channel (transparency) without the artifacts introduced
* by <code>scaleInto()</code>.
*
* @param bmpSrc Source Bitmap
* @param nWidth New Width
* @param nHeight New Height
* @param nFilterType Filter quality to use. Can be <code>Bitmap.FILTER_LANCZOS</code>,
* <code>Bitmap.FILTER_BILINEAR</code> or
* <code>Bitmap.FILTER_BOX</code>.
* @param nAspectRatio Specifies how the picture is resized. Can be
* <code>Bitmap.SCALE_TO_FIT</code>,
* <code>Bitmap.SCALE_TO_FILL</code> or
* <code>Bitmap.SCALE_STRETCH</code>.
* @return The resized Bitmap in a new object.
*/
public static Bitmap ResizeTransparentBitmap(Bitmap bmpSrc, int nWidth, int nHeight, int nFilterType, int nAspectRatio)
{
if(bmpsrc== null)
return null;
//Get the original dimensions of the bitmap
int nOriginWidth = bmpSrc.getWidth();
int nOriginHeight = bmpSrc.getHeight();
if(nWidth == nOriginWidth && nHeight == nOriginHeight)
return bmpSrc;
//Prepare a drawing bitmap and graphic object
Bitmap bmpOrigin = new Bitmap(nOriginWidth, nOriginHeight);
Graphics graph = Graphics.create(bmpOrigin);
//Create a line of transparent pixels for later use
int[] aEmptyLine = new int[nWidth];
for(int x = 0; x < nWidth; x++)
aEmptyLine[x] = 0x00000000;
//Create two scaled bitmaps
Bitmap[] bmpScaled = new Bitmap[2];
for(int i = 0; i < 2; i++)
{
//Draw the bitmap on a white background first, then on a black background
graph.setColor((i == 0) ? Color.WHITE : Color.BLACK);
graph.fillRect(0, 0, nOriginWidth, nOriginHeight);
graph.drawBitmap(0, 0, nOriginWidth, nOriginHeight, bmpSrc, 0, 0);
//Create a new bitmap with the desired size
bmpScaled[i] = new Bitmap(nWidth, nHeight);
if(nAspectRatio == Bitmap.SCALE_TO_FIT)
{
//Set the alpha channel of all pixels to 0 to ensure transparency is
//applied around the picture, if needed by the transformation
for(int y = 0; y < nHeight; y++)
bmpScaled[i].setARGB(aEmptyLine, 0, nWidth, 0, y, nWidth, 1);
}
//Scale the bitmap
bmpOrigin.scaleInto(bmpScaled[i], nFilterType, nAspectRatio);
}
//Prepare objects for final iteration
Bitmap bmpFinal = bmpScaled[0];
int[][] aPixelLine = new int[2][nWidth];
//Iterate every line of the two scaled bitmaps
for(int y = 0; y < nHeight; y++)
{
bmpScaled[0].getARGB(aPixelLine[0], 0, nWidth, 0, y, nWidth, 1);
bmpScaled[1].getARGB(aPixelLine[1], 0, nWidth, 0, y, nWidth, 1);
//Check every pixel one by one
for(int x = 0; x < nWidth; x++)
{
//If the pixel was untouched (alpha channel still at 0), keep it transparent
if(((aPixelLine[0][x] >> 24) & 0xff) == 0)
aPixelLine[0][x] = 0x00000000;
else
{
//Compute the alpha value based on the difference of intensity
//in the red channel
int nAlpha = ((aPixelLine[1][x] >> 16) & 0xff) -
((aPixelLine[0][x] >> 16) & 0xff) + 255;
if(nAlpha == 0)
aPixelLine[0][x] = 0x00000000; //Completely transparent
else if(nAlpha >= 255)
aPixelLine[0][x] |= 0xff000000; //Completely opaque
else
{
//Compute the value of the each channel one by one
int nRed = ((aPixelLine[0][x] >> 16 ) & 0xff);
int nGreen = ((aPixelLine[0][x] >> 8 ) & 0xff);
int nBlue = (aPixelLine[0][x] & 0xff);
nRed = (int)(255 + (255.0 * ((double)(nRed-255)/(double)nAlpha)));
nGreen = (int)(255 + (255.0 * ((double)(nGreen-255)/(double)nAlpha)));
nBlue = (int)(255 + (255.0 * ((double)(nBlue-255)/(double)nAlpha)));
if(nRed < 0) nRed = 0;
if(nGreen < 0) nGreen = 0;
if(nBlue < 0) nBlue = 0;
aPixelLine[0][x] = nBlue | (nGreen<<8) | (nRed<<16) | (nAlpha<<24);
}
}
}
//Change the pixels of this line to their final value
bmpFinal.setARGB(aPixelLine[0], 0, nWidth, 0, y, nWidth, 1);
}
return bmpFinal;
}
}
Here’s an example of how to call the function:
Bitmap bmp = Bitmap.getBitmapResource("picture.png");
Bitmap bmpResized = GPATools.ResizeTransparentBitmap(bmp, 30, 60,
Bitmap.FILTER_LANCZOS, Bitmap.SCALE_TO_FIT);
Here’s the result:



It’s really inspiring that when the predefined function didn’t work as you expected, you just made one yourself that did a great work.
I’m starting in programming, and at least at first I would have become just really frustrated.
Thanks for sharing
The results are excellent!