在Java中按比例缩小后，为什么此图像看起来如此糟糕？

Question

Here is the original image:
http://rank.my/public/images/uploaded/orig-4193395691714613396.png

And here it is scaled down to 300x225:

http://rank.my/public/images/uploaded/norm-4193395691714613396.png

And here it is scaled down to 150x112:

http://rank.my/public/images/uploaded/small-4193395691714613396.png

As you can see, 300x225 looks pretty bad, and 150x112 looks awful. Here is the
code I’m using to scale down the image:

private static BufferedImage createResizedCopy(final BufferedImage source, final int destWidth,
        final int destHeight) {
    final BufferedImage resized = new BufferedImage(destWidth, destHeight, source.getType());
    final Graphics2D bg = resized.createGraphics();
    bg.setRenderingHint(RenderingHints.KEY_INTERPOLATION, RenderingHints.VALUE_INTERPOLATION_BILINEAR);
    bg.setRenderingHint(RenderingHints.KEY_RENDERING, RenderingHints.VALUE_RENDER_QUALITY);
    bg.setRenderingHint(RenderingHints.KEY_ANTIALIASING, RenderingHints.VALUE_ANTIALIAS_ON);
    final float sx = (float) destWidth / source.getWidth();
    final float sy = (float) destHeight / source.getHeight();
    bg.scale(sx, sy);
    bg.drawImage(source, 0, 0, null);
    bg.dispose();
    return resized;
}

我在这里做错了什么？图像缩放不必特别
快，质量绝对是优先于速度的。我使用了错误的
技术吗？

Answer 1

有三种方法可以解决缩小问题。首先是分
多个步骤进行，每个步骤最多减少75％。第二种是
在调整大小之前模糊图像。缩小程度越多
，模糊程度就越高。第三种方法是使用一种方法，该方法使用
朴素的双线性或双三次插值方法使用的2x2至4x4像素块进行滤波。
随着收缩因子的增大，
滤镜使用的像素块也应增大，否则您将看到混叠伪像，如此处所示。