Update PDF image in-place

Question

I am trying to replace an image stream within an SDF document, using PDFNet 7.0.4 and netcoreapp3.1. As much as possible, I want to maintain the original object and its metadata; same dimensions, color system, compression, etc. Ideally object number and even generation would remain the same as well - the goal is that a before and after comparison would show only the changed pixels within the stream.

I'm getting the raw pixel data as a Stream object with this method:

private Stream GetImageData(int objectNum)
{
    var image = new PDF.Image(sdfDoc.GetObj(objectNum));

    var bits = image.GetBitsPerComponent();
    var channels = image.GetComponentNum();
    var bytesPerChannel = bits / 8;
    var height = image.GetImageHeight();
    var width = image.GetImageWidth();

    var data = image.GetImageData();
    var len = height * width * channels * bytesPerChannel;

    using (var reader = new pdftron.Filters.FilterReader(data))
    {
        var buffer = new byte[len];
        reader.Read(buffer);                

        return new MemoryStream(buffer);
    }
}

After manipulating the image data, I want to update it before saving the underlying SDFDoc object. I've tried using the following method:

private void SetImageData(int objectNum, Stream stream)
{
    var image = new PDF.Image(sdfDoc.GetObj(objectNum));

    var bits = image.GetBitsPerComponent();
    var channels = image.GetComponentNum();
    var bytesPerChannel = bits / 8;
    var height = image.GetImageHeight();
    var width = image.GetImageWidth();

    var len = height * width * channels * bytesPerChannel;
    if (stream.Length != len) { throw new DataMisalignedException("Stream length does not match expected image dimensions"); }

    using (var ms = new MemoryStream())
    using (var writer = new pdftron.Filters.FilterWriter(image.GetImageData()))
    {
        stream.CopyTo(ms);
        writer.WriteBuffer(ms.ToArray());
    }
}

This runs without error, but nothing actually appears to get updated. I've tried playing around with SDFObj.SetStreamData(), but haven't been able to make that work either. What is the lowest impact, highest performance way to directly replace just the raw pixel data within an image stream?

---

edit

I have this halfway working with this method:

private void SetImageData(int objectNum, Stream stream)
{
    var sdfObj = sdfDoc.GetObj(objectNum);
    var image = new PDF.Image(sdfObj);

    var bits = image.GetBitsPerComponent();
    var channels = image.GetComponentNum();
    var bytesPerChannel = bits / 8;
    var height = image.GetImageHeight();
    var width = image.GetImageWidth();

    var len = height * width * channels * bytesPerChannel;
    if (stream.Length != len) { throw new DataMisalignedException("Stream length does not match expected image dimensions"); }

    var buffer = new byte[len];
    stream.Read(buffer, 0, len);
    sdfObj.SetStreamData(buffer);
    sdfObj.Erase("Filters");
}

This works as expected, but with the obvious caveat that it just ignores any existing compression and turns the image into a raw uncompressed stream.

I've tried sdfObj.SetStreamData(buffer, image.GetImageData()); and sdfObj.SetStreamData(buffer, image.GetImageData().GetAttachedFilter()); and this does update the object in the file, but the resulting image fails to render.

Answer 1

The following code shows how to retain an Image object, but change the actual stream data.

static private Stream GetImageData(Obj o)
{
    var image = new pdftron.PDF.Image(o);

    var bits = image.GetBitsPerComponent();
    var channels = image.GetComponentNum();
    var bytesPerChannel = bits / 8;
    var height = image.GetImageHeight();
    var width = image.GetImageWidth();

    var data = image.GetImageData();
    var len = height * width * channels * bytesPerChannel;

    using (var reader = new pdftron.Filters.FilterReader(data))
    {
        var buffer = new byte[len];
        reader.Read(buffer);
        return new MemoryStream(buffer);
    }
}

static private void SetImageData(PDFDoc doc, Obj o, Stream stream)
{

    var image = new pdftron.PDF.Image(o);

    var bits = image.GetBitsPerComponent();
    var channels = image.GetComponentNum();
    var bytesPerChannel = bits / 8;
    var height = image.GetImageHeight();
    var width = image.GetImageWidth();

    var len = height * width * channels * bytesPerChannel;
    if (stream.Length != len) { throw new DataMisalignedException("Stream length does not match expected image dimensions"); }

    o.Erase("DecodeParms"); // Important: this won'be accurate after SetStreamData
    // now we actually do the stream swap
    o.SetStreamData((stream as MemoryStream).ToArray(), new FlateEncode(null));
}

static private void InvertPixels(Stream stream)
{
    // This function is for DEMO purposes
    // this code assumes 3 channel 8bit
    long length = stream.Length;
    long pixels = length / 3;
    for(int p = 0; p < pixels; ++p)
    {
        int c1 = stream.ReadByte();
        int c2 = stream.ReadByte();
        int c3 = stream.ReadByte();
        stream.Seek(-3, SeekOrigin.Current);
        stream.WriteByte((byte)(255 - c1));
        stream.WriteByte((byte)(255 - c2));
        stream.WriteByte((byte)(255 - c3));
    }
    stream.Seek(0, SeekOrigin.Begin);
}

And then here is sample code to use.

static void Main(string[] args)
{
    PDFNet.Initialize();

    var x = new PDFDoc(@"2002.04610.pdf");
    x.InitSecurityHandler();

    var o = x.GetSDFDoc().GetObj(381);
    Stream source = GetImageData(o);
    InvertPixels(source);
    SetImageData(x, o, source);
    x.Save(@"2002.04610-MOD.pdf", SDFDoc.SaveOptions.e_remove_unused);
}