need to combine several jpegs into a bigger jpeg in a lossless way programmaticaly (.net)

Question

i have several jpeg images and i want to combine them into one big jpeg image.
i can do that by creating a Bitmap and then combining them there but that way if i save it again as jpeg the image will deteriorate.
so is there any method that i can do that without losing quality from decoding/encoding?
in acdsee program i saw an option to rotate jpegs images without deteriorating quality.
so there should be a way to combine several images without losing quality.
thanks

Answer 1

According to Wikipedia/Jpeg it could be possible if your images have sizes that are multiples of 16.

Wikipedia/Lossless editing/JPEG also talks about JPEGJoin that can combine several images.

There is nothing build in in .NET Framework, but you might be able to use the above tools from C#.

Answer 2

Almost all lossless JPEG tools are based on jpegtran from http://sylvana.net/jpegcrop/jpegtran/ (source code is available).

What you need is to extend the jpeg canvas and then use the still experimental "drop" functionality to put an image into another image.

I am not sure of this, but I think the images need to use the same quantization tables (~ encoding quality) in order to be joined losslessly.

Answer 3

well i have written the code and so i wanted to share it here:)
please note that the code wont work on all situations but its fine for my use.
i am using the LibJpeg.Net library http://bitmiracle.com/libjpeg .
also there is a bug in the library (or a bug in me:) ) that you cant get the component height_in_blocks this is why that code will only work on square tiles.
i think that the images need to have the same quantization table as Vlasta mentioned.
i think this code can be expanded to support that but well i didnt need such support.
and now here comes the code :)

public void CreateBigImage()
{
    const int iTileWidth = 256;
    const int iTileHeigth = 256;
    int iImageWidthInTiles = 2;
    int iImageHeigthInTiles = 2;

    //Open Image to read its header in the new image
    BitMiracle.LibJpeg.Classic.jpeg_decompress_struct objJpegDecompressHeader = new BitMiracle.LibJpeg.Classic.jpeg_decompress_struct();
    System.IO.FileStream objFileStreamHeaderImage = new System.IO.FileStream(GetImagePath(0), System.IO.FileMode.Open, System.IO.FileAccess.Read);
    objJpegDecompressHeader.jpeg_stdio_src(objFileStreamHeaderImage);
    objJpegDecompressHeader.jpeg_read_header(true);

    BitMiracle.LibJpeg.Classic.jvirt_array<BitMiracle.LibJpeg.Classic.JBLOCK>[] varrJBlockBigImage = new BitMiracle.LibJpeg.Classic.jvirt_array<BitMiracle.LibJpeg.Classic.JBLOCK>[10];
    for (int i = 0; i < 3; i++)//3 compounds per image (YCbCr)
    {
        int iComponentWidthInBlocks = objJpegDecompressHeader.Comp_info[i].Width_in_blocks;
        int iComponentHeigthInBlocks = iComponentWidthInBlocks;//there is no Height_in_blocks in the library so will use widht for heigth also (wont work if image is not rectangular)
        varrJBlockBigImage[i] = BitMiracle.LibJpeg.Classic.jpeg_common_struct.CreateBlocksArray(iComponentWidthInBlocks * iImageWidthInTiles, iComponentHeigthInBlocks * iImageHeigthInTiles);
    }

    for (int iX = 0; iX < iImageWidthInTiles; iX++)
    {
        for (int iY = 0; iY < iImageHeigthInTiles; iY++)
        {
            WriteImageToJBlockArr(varrJBlockBigImage, GetImagePath(iY*iImageHeigthInTiles+iX), iX, iY);
        }
    }

    System.IO.FileStream objFileStreamMegaMap = System.IO.File.Create(GetImagePath(999));
    BitMiracle.LibJpeg.Classic.jpeg_compress_struct objJpegCompress = new BitMiracle.LibJpeg.Classic.jpeg_compress_struct();
    objJpegCompress.jpeg_stdio_dest(objFileStreamMegaMap);
    objJpegDecompressHeader.jpeg_copy_critical_parameters(objJpegCompress);//will copy the critical parameter from the header image
    objJpegCompress.Image_height = iTileHeigth * iImageHeigthInTiles;
    objJpegCompress.Image_width = iTileWidth * iImageWidthInTiles;
    objJpegCompress.jpeg_write_coefficients(varrJBlockBigImage);
    objJpegCompress.jpeg_finish_compress();
    objFileStreamMegaMap.Close();

    objJpegDecompressHeader.jpeg_abort_decompress();
    objFileStreamHeaderImage.Close();
}

public void WriteImageToJBlockArr(BitMiracle.LibJpeg.Classic.jvirt_array<BitMiracle.LibJpeg.Classic.JBLOCK>[] varrJBlockNew, string strImagePath, int iTileX, int iTileY)
{
    BitMiracle.LibJpeg.Classic.jpeg_decompress_struct objJpegDecompress = new BitMiracle.LibJpeg.Classic.jpeg_decompress_struct();
    System.IO.FileStream objFileStreamImage = new System.IO.FileStream(strImagePath, System.IO.FileMode.Open, System.IO.FileAccess.Read);
    objJpegDecompress.jpeg_stdio_src(objFileStreamImage);
    objJpegDecompress.jpeg_read_header(true);
    BitMiracle.LibJpeg.Classic.jvirt_array<BitMiracle.LibJpeg.Classic.JBLOCK>[] varrJBlockOrg = objJpegDecompress.jpeg_read_coefficients();
    for (int i = 0; i < 3; i++)//3 compounds per image (YCbCr)
    {
        int iComponentWidthInBlocks = objJpegDecompress.Comp_info[i].Width_in_blocks;
        int iComponentHeigthInBlocks = iComponentWidthInBlocks;//there is no Height_in_blocks in the library so will use widht for heigth also (wont work if image is not rectangular)
        for (int iY = 0; iY < iComponentHeigthInBlocks; iY++)
        {
            for (int iX = 0; iX < iComponentWidthInBlocks; iX++)
            {
                varrJBlockNew[i].Access(iY + iTileY * iComponentHeigthInBlocks, 1)[0][iX + iTileX * iComponentWidthInBlocks] = varrJBlockOrg[i].Access(iY, 1)[0][iX];
            }
        }
    }
    objJpegDecompress.jpeg_finish_decompress();
    objFileStreamImage.Close();
}