You can always resize images with any software, but some developers need to create softwares like piknic! This is what’s this tip is all about…

I’ve given the above algorithm to one of our developers

BufferedImage temp = javax.imageio.ImageIO.read(new File(imagefile));
/* determine thumbnail size from WIDTH and HEIGHT */
int thumbWidth = width;
int thumbHeight = height;
int imageWidth = temp.getWidth(null);
int imageHeight = temp.getHeight(null);

int tempWidth;
int tempHeight;
int x = 0;
int y = 0;

if (imageWidth < imageHeight) {
tempWidth = width;
tempHeight = (int)(((double)imageHeight*width)/imageWidth);
y = -(tempHeight – tempWidth)/2;
}
else {
tempHeight = height;
tempWidth = (int)(((double)imageWidth*height)/imageHeight);
x = -(tempWidth – tempHeight)/2;
}
Image temp1 = temp.getScaledInstance(tempWidth, tempHeight, Image.SCALE_SMOOTH);
image = new BufferedImage(width, height, BufferedImage.TYPE_INT_RGB);
image.getGraphics().drawImage(temp1, x, y, null);
try {
javax.imageio.ImageIO.write(image, “png”, new File(“output.png”));
} catch (IOException e) {
e.printStackTrace();
}

But he says

but the ImageIO does not recognize a number of images uploaded by the dealer. It seems to have problems with non-standard colour palettes. This is something we saw frequently in production untill we switched to the current method with the toolkit.

After reading many documents (althought only one gave a good clue to fix it), I came to the conclusion that the Java 2D library doesn’t work well when you down-resize from very distant sizes. Aproximately when the scale proportion is below the 50% of the image size. If you apply a progressive resizing (by steps), you get a nice and smooth result. I made a recursive function and it worked.

BufferedImage source, target…;

double scalex = (double) target.getWidth() / source.getWidth();
double scaley = (double) target.getHeight() / source.getHeight();
AffineTransform at = AffineTransform.getScaleInstance(scalex, scaley);

AffineTransformOp affineTransformOp = new AffineTransformOp(at, AffineTransformOp.TYPE_BICUBIC);

affineTransformOp.filter(source, target);