Rendering of an 8Bit-tiff with java.awt produces crap

Question

Hi Folks,
i do a downsampling of tiff images. The source tiffs are colored 8 Bit greyscale with a 256 color-table. Now i have 2 possibilities to resample with java.awt:
1. the target-tiff is of 24Bit true color:
BufferedImage objDownsample = new BufferedImage(iTargetWidth, iTargetHeight, BufferedImage.TYPE_INT_RGB);The result is a perfect looking, but oversized tiff with 16 million colors each pixel.
2. the target-tiff is of 8Bit like the source-tiff:
BufferedImage objDownsample = new BufferedImage(iTargetWidth, iTargetHeight, BufferedImage.TYPE_BYTE_INDEXED,(IndexColorModel)image.getColorModel());The result is a small sized tiff image of 8Bit. Problem: it now has visible vertical stripes of two colors (which both composed result the source-color i assume).
Does anybody know what's wrong here and how to retrieve the 8Bit color-image without stripes?
Here comes the whole source:
    private BufferedImage resize(int newHeight, int newWidth, BufferedImage image) {
        int iTargetHeight = newHeight;
        int iTargetWidth = newWidth;
        //Create a BufferedImage that fits
        BufferedImage objDownsample = new BufferedImage(iTargetWidth, iTargetHeight, BufferedImage.TYPE_BYTE_INDEXED,(IndexColorModel)image.getColorModel());
        //A map with all necessary rendering hints to optimize the quality of the image
        Map<java.awt.RenderingHints.Key, java.lang.Object> obj_Map = new HashMap<java.awt.RenderingHints.Key, java.lang.Object>();       
        obj_Map.put(RenderingHints.KEY_DITHERING, RenderingHints.VALUE_DITHER_ENABLE);
        obj_Map.put(RenderingHints.KEY_ANTIALIASING, RenderingHints.VALUE_ANTIALIAS_ON);
        obj_Map.put(RenderingHints.KEY_INTERPOLATION, RenderingHints.VALUE_INTERPOLATION_BILINEAR);
      // Draw the scaled image
      Graphics2D graphics2D = objDownsample.createGraphics();
        graphics2D.addRenderingHints(obj_Map);
        graphics2D.drawImage(image, 0, 0, iTargetWidth, iTargetHeight, null);
        return objDownsample;
    }Thanks
Albrecht

Answer 1

As far as I can tell, this solution only allows compositing within the component that is currently being painted. What would be nice is if I could add a custom composite to say .. a JLabel, which paints based on the colors of the underlying pixel data (ie. from its parent container). In this way you could make a JLabel that always inverts the color of what's underneath.
Here, painting to the screen using a composite allows me to get the pixels underneath as the destination raster, the pixels in the JLabel as the source raster, and what actually gets painted which is the destination output raster.
If I paint to a BufferedImage instead, the composite's 'destination raster' is the new BufferedImage's default values which are all black. Thus compositing over this would not have the desired effects described above (inverting the underlying image for instance). The missing piece when rendering to the BufferedImage is first copying the underlying screen data into a BufferedImage before painting to it. How do I do that? Something tells me it's very difficult.