感谢Rob的回答,我们现在知道为什么颜色混乱了.

问题是双重的：

> ImageIO用于编写JPEG的默认JPEGImageWriter,不会以其他软件理解的方式编写带有alpha的JPEG(这是一个已知问题).

>当将null作为目标传递给ResampleOp.filter(src,dest)并且过滤器方法是FILTER_TRIANGLE时,将创建一个新的BufferedImage,其中包含alpha(实际上是BufferedImage.TYPE_INT_ARGB).

重新采样后剥离alpha将起作用.但是,还有另一种方法可能更快并节省一些内存.也就是说,不是传递空目标,而是传递适当大小和类型的BufferedImage：

public static void main(String[] args) throws IOException {

// Read input

File input = new File(args[0]);

BufferedImage inputImage = ImageIO.read(input);

// Make any transparent parts white

if (inputImage.getTransparency() == Transparency.TRANSLUCENT) {

// NOTE: For BITMASK images, the color model is likely IndexColorModel,

// and this model will contain the "real" color of the transparent parts

// which is likely a better fit than unconditionally setting it to white.

// Fill background with white

Graphics2D graphics = inputImage.createGraphics();

try {

graphics.setComposite(AlphaComposite.DstOver); // Set composite rules to paint "behind"

graphics.setPaint(Color.WHITE);

graphics.fillRect(0, 0, inputImage.getWidth(), inputImage.getHeight());

}

finally {

graphics.dispose();

}

}

// Resample to fixed size

int width = 100;

int height = 100;

BufferedImageOp resampler = new ResampleOp(width, height, ResampleOp.FILTER_TRIANGLE);

// Using explicit destination, resizedImg will be of TYPE_INT_RGB

BufferedImage resizedImg = resampler.filter(inputImage, new BufferedImage(width, height, BufferedImage.TYPE_INT_RGB));

// Write output as JPEG

ImageIO.write(resizedImg, "JPEG", new File(input.getParent(), input.getName().replace('.', '_') + ".jpg"));

}