调用Raster.setRect()时的ArrayIndexOutOfBoundsException

Question

我写了一个读取图像的方法。我循环遍历图像文件列表并读取它们。在一些迭代之后，当调用Raster.setRect(栅格)方法时出现ArrayIndexOutOfBoundsException。图像的尺寸和边界都很好-请参阅stacktrace。

public static BufferedImage readImage(File imageFile) {

    // Find a suitable ImageReader
    Iterator<ImageReader> readers = ImageIO.getImageReadersBySuffix("jpg");
    ImageReader reader = null;
    while (readers.hasNext()) {
        reader = (ImageReader) readers.next();
        if (reader.canReadRaster()) {
            break;
        }
    }

    // Stream the image file (the original CMYK image)
    ImageInputStream input = null;
    try {
        input = ImageIO.createImageInputStream(imageFile);
    } catch (IOException e) {
        logger.error("Error creating InputStream on File {}", imageFile
                .getName());
        e.printStackTrace();
    }
    reader.setInput(input);

    // Read the image raster
    Raster raster = null;
    try {
        raster = reader.readRaster(0, null);
    } catch (IOException e) {
        logger
                .error("Error reading Raster of file {}", imageFile
                        .getName());
        e.printStackTrace();
    }

    // Create a new RGB image
    BufferedImage bi = new BufferedImage(raster.getWidth(), raster
            .getHeight(), BufferedImage.TYPE_INT_RGB);

    // Fill the new image with the old raster
    logger.debug("Height {} and width {} of original raster", raster
            .getHeight(), raster.getWidth());
    logger.debug("Height {} and width {} of new raster", bi.getRaster()
            .getHeight(), bi.getRaster().getWidth());
    logger.debug("NumBands original raster {}", bi.getRaster().getBounds());
    logger.debug("NumBands new raster {}", bi.getRaster().getBounds());
    bi.getRaster().setRect(raster);

    // Close and flush the reader
    try {
        input.close();
    } catch (IOException e) {
        logger.error("Error closing the reader for file {}", imageFile
                .getName());
        e.printStackTrace();
    }

    return bi;
}

下面是堆栈跟踪：

Exception in thread "main" java.lang.ArrayIndexOutOfBoundsException: 310
at java.awt.image.SinglePixelPackedSampleModel.setPixels(SinglePixelPackedSampleModel.java:689)
at java.awt.image.WritableRaster.setPixels(WritableRaster.java:565)
at java.awt.image.WritableRaster.setRect(WritableRaster.java:467)
at java.awt.image.WritableRaster.setRect(WritableRaster.java:403)
at .. .utils.ImageUtils.readImage(ImageUtils.java:228)
at .. .image.scripts.ImageMatchingScript.main(ImageMatchingScript.java:86)

所有图像都具有相同的尺寸371x310

下面是我打印的调试信息：

13:53:17.065 [main] DEBUG .. .ImageUtils - Height 371 and width 310 of original raster
13:53:17.065 [main] DEBUG .. .ImageUtils - Height 371 and width 310 of new raster
13:53:17.065 [main] DEBUG .. .ImageUtils - NumBands original raster java.awt.Rectangle[x=0,y=0,width=310,height=371]
13:53:17.065 [main] DEBUG .. .ImageUtils - NumBands new raster java.awt.Rectangle[x=0,y=0,width=310,height=371]

文档说: ArrayIndexOutOfBoundsException -如果坐标不在边界内，或者如果fArray太小而无法容纳输入。

对我来说，尺寸/边界似乎很好，也许这是fArray的问题？

Answer 1

我想问题不在于图像的尺寸，而在于光栅中的波段数。您可以创建自己的BufferedImage.TYPE_INT_RGB。可能失败的图像是带有alpha通道的PNG，在这种情况下，光栅有一个额外的波段。

我想知道为什么您要以这种复杂的方式进行加载，为什么不直接使用ImageIo.read(InputStream)呢？这将给你一个BufferedImage，不需要自己转换它。

编辑:尝试此方法读取并转换为RGB：

InputStream input = null;
try {
input = new FileInputStream(...); // whatever your source is
    BufferedImage rawImage = ImageIO.read(input);
    if (rawImage == null)
        return null;
    BufferedImage rgbImage = new BufferedImage(rawImage.getWidth(null), rawImage.getHeight(null), BufferedImage.TYPE_INT_RGB);
    Graphics g = rgbImage.createGraphics();
    // if rawImage contains alpha, consider filling the rgbImage with a default BG color here
    g.drawImage(rawImage, 0, 0, null);
    g.dispose();
    rawImage.flush();
    return rgbImage;
} catch (Exception e) {
// handle error
    return null;
} finally {
    // close input if open
    if (input != null) {
        try {
            input.close();
        } catch (IOException e) {}
    }
}

只是把代码拼凑在一起，可能包含打字错误。这样你就可以得到RGB图像，而不管源文件是什么，甚至无需考虑源图像属性。