Re: [Gdcm2] Buffer problem in Qt environment

[Sylvain A. <syl...@ho...>]

Okay, here's the whole source code used for the conversion:

============================================================================
gdcm::ImageReader ir;
    ir.SetFileName("wedges.ima");

    if(!ir.Read())
    {
                    //Read failed
    }


    std::cout<<"Getting image from ImageReader..."<<std::endl;


    gdcm::Image &gimage = ir.GetImage();


    const unsigned int* dimension = gimage.GetDimensions();


    int dimX = int(dimension[0]);
    int dimY = int(dimension[1]);
//This buffer has been declared elsewhere


    buffer = new char[gimage.GetBufferLength()];


    gimage.GetBuffer(buffer);


    //Let's now convert each 16-bits pixel to a 8-bits pixel, leveled to the maximum intensity value of the image
    //Because the image is a 16-bits grayscale, we thus have 65536 available values to store the gray value.
    // BUT, usually, the value doesn't oversize around 2000~2200. So if we directly convert the 16-bits to 8-bits (from max. 65536 to max255)
    // and try to display it, we will have very dark images. Then, we will level all the intensities to the maximum intensity value.

    //We are gonna use a RGB888 format for Qt display. In a grayscale image, R=G=B = the gray value.
    //The buffer length, in char, will thus be: 256*256*3 (if the image is 256x256) = 196608 chars.


    buffer16 = new unsigned short int[dimX*dimY];
    maxValue = 0;

    for(unsigned int i = 0; i < gimage.GetBufferLength(); i+=2)
    {

                    //@todo: think about #typedef to avoid these long conversions lines.
                    //Bitwise left shift and bitwise OR
                    unsigned short int pixel16bitValue = ((((unsigned short int) ((unsigned char) buffer [i+1]))<<8)|(unsigned char) buffer [i]);

                    buffer16[i/2] = pixel16bitValue;

            if(pixel16bitValue > maxValue)
            {
                    maxValue = pixel16bitValue;
            }
    }


    unsigned char *buffer888 = new unsigned char[dimX*dimY*3];


    for(int i = 0; i < dimX*dimY; i++)
    {
            unsigned char pixel8bitValue = (unsigned char) (buffer16[i]*255/maxValue);
            buffer888[i*3] = pixel8bitValue;
            buffer888[i*3+1] = pixel8bitValue;
            buffer888[i*3+2] = pixel8bitValue;
    }


    QImage *imageQt = new QImage(buffer888, dimX, dimY, QImage::Format_RGB888);



    //The DICOM image is a 16-bits image
    //For a 256x256 image, it's 256x256 = 65536 pixels.
    //So, normally in a char* (1 char = 1 byte = 1 bit), it does have a length of 65536    //BUT as it's a 16-bits image (2*8-bits), we thus have 131072 chars
============================================================================
Hope this will help other people to get everything right!
Haven't tried the PGM yet, but I guess that's easily comparable.
By the way, I've tried accessing your wiki page, but nothing displays for now?
And should I try to post a new bug for this memory problem I evoked yesterday?

Sylvain


> Date: Wed, 17 Jun 2009 13:41:14 +0200
> Subject: Re: [Gdcm2] Buffer problem in Qt environment
> From: mat...@gm...
> To: syl...@ho...
> CC: gdc...@li...
> 
> Hi Sylvain,
> 
>   congrats ! I realize this was my mistake, I quickly browse through
> the documentation of the image format:
> 
> http://doc.trolltech.com/4.5/qimage.html#Format-enum
> 
>   and I thought QImage could handle 16bits monochrome image, while it does not.
> 
>   Please share your conversion code and I'll put it in Examples/Cxx
> (did you use C++ ?).
> 
>   As a side note, you can use the command line gdcmimg to convert a
> DICOM image to pgm, and then you could load the QImage using the load
> API:
> 
> http://doc.trolltech.com/4.5/qimage.html#reading-and-writing-image-files
> 
>   so that you can check your conversion from 8bits is identical to
> what 'load()' would do with an input 16bits file (I am surprise there
> is no such filter already in Qt).
> 
> Eg:
> 
>   $ gdcmconv --raw  gdcmData/012345.002.050.dcm /tmp/dummy.dcm
>   $ gdcmimg /tmp/dummy.dcm /tmp/dummy.pgm
>   $ identify /tmp/dummy.pgm
> 
> 
>         ~
>   /tmp/dummy.pgm PNM 256x256 256x256+0+0 PseudoClass 65536c 16-bit 128.016kb
> 
> thanks for your contribution !
> 
> 
> 
> -- 
> Mathieu

_________________________________________________________________
Téléphonez gratuitement à tous vos proches avec Windows Live Messenger  !  Téléchargez-le maintenant ! 
http://www.windowslive.fr/messenger/1.asp