Choosing represenation problem.

[zaq]

Hi,

I write my own Store SCU, which operates only on memory located datasets and recently I ecnoutered a problem while trying to use compressed transfer syntaxes, i.e. the JPEG ones.

Here's the code that causes the problem:

Code: Select all

	DcmDataset * dcmDataset = new DcmDataset();

	dcmDataset->putAndInsertUint16( DCM_FileMetaInformationVersion, 0x0001 );
	dcmDataset->putAndInsertString( DCM_SOPClassUID,                UID_SecondaryCaptureImageStorage );
	dcmDataset->putAndInsertString( DCM_SOPInstanceUID,             dcmGenerateUniqueIdentifier( uid, SITE_INSTANCE_UID_ROOT ) );
	dcmDataset->putAndInsertString( DCM_PhotometricInterpretation,  "RGB" ); 
	dcmDataset->putAndInsertUint16( DCM_SamplesPerPixel,            3 );
	dcmDataset->putAndInsertString( DCM_NumberOfFrames,             "1" );
	dcmDataset->putAndInsertUint16( DCM_BitsAllocated,              8 );
	dcmDataset->putAndInsertUint16( DCM_BitsStored,                 7 ); 
	dcmDataset->putAndInsertUint16( DCM_HighBit,                    7 );
	dcmDataset->putAndInsertUint16( DCM_PixelRepresentation,        0 );
	dcmDataset->putAndInsertUint16( DCM_PlanarConfiguration,        0 );
	dcmDataset->putAndInsertUint16( DCM_SmallestImagePixelValue,    0 );
	dcmDataset->putAndInsertUint16( DCM_LargestImagePixelValue,     255 );
	dcmDataset->putAndInsertUint16( DCM_Rows,                       image.height() );
	dcmDataset->putAndInsertUint16( DCM_Columns,                    image.width() );


	unsigned char * data = new unsigned char [ image.height() * image.width() * 3 ];

	// Processing image pixels.
	// ...
	dcmDataset->putAndInsertUint8Array(
		DCM_PixelData, data, image.height() * image.width() * 3
	);

	DJ_RPLossy rp_lossy;
	DJEncoderRegistration::registerCodecs();
	if ( dcmDataset->chooseRepresentation( EXS_JPEGProcess1TransferSyntax, & rplossy ).bad() ) {
		// Handle error.
	}
	if ( ! dcmDataset->canWriteXfer( EXS_JPEGProcess1TransferSyntax ) ) {
		// Handle error.
	}

As you can see I create a blank DcmDataset, then insert to it some Image Module tags and then try to change its representation for the compressed one. Unfortunetely `chooseRepresentation()' method returns bad status.

I used debugger to find out what's wrong and stepped into the method. It appears that the `DcmImage::Init()' method (which is called when creating temporary object) doesn't support RGB Photomoetric Interpretation. Why is that?


The full call stack is as follows:

Code: Select all

	theapp.exe!DicomImage::Init()  Line 197	C++
 	theapp.exe!DicomImage::DicomImage(DcmObject * object=0x023dd218, E_TransferSyntax xfer=EXS_LittleEndianImplicit, const unsigned long flags=0, const unsigned long fstart=0, const unsigned long fcount=0)  Line 105	C++
 	theapp.exe!DJCodecEncoder::encodeColorImage(bool YBRmode=false, DcmItem * dataset=0x023dd218, const DcmRepresentationParameter * toRepParam=0x0012d81c, DcmPixelSequence * & pixSeq=0x00000000, const DJCodecParameter * cp=0x023dd910, double & compressionRatio=0.00000000000000000)  Line 286 + 0x36 bytes	C++
 	theapp.exe!DJCodecEncoder::encode(const unsigned short * __formal=0x02ed0040, const unsigned short * __formal=0x02ed0040, const DcmRepresentationParameter * toRepParam=0x0012d81c, DcmPixelSequence * & pixSeq=0x00000000, const DcmCodecParameter * cp=0x023dd910, DcmStack & objStack={...})  Line 155 + 0x27 bytes	C++
 	theapp.exe!DcmCodecList::encode(E_TransferSyntax fromRepType=EXS_LittleEndianExplicit, const unsigned short * pixelData=0x02ed0040, const unsigned long length=2805000, E_TransferSyntax toRepType=EXS_JPEGProcess1TransferSyntax, const DcmRepresentationParameter * toRepParam=0x0012d81c, DcmPixelSequence * & toPixSeq=0x00000000, DcmStack & pixelStack={...})  Line 449 + 0x43 bytes	C++
 	theapp.exe!DcmPixelData::encode(const DcmXfer & fromType={...}, const DcmRepresentationParameter * fromParam=0x00000000, DcmPixelSequence * fromPixSeq=0x00000000, const DcmXfer & toType={...}, const DcmRepresentationParameter * toParam=0x0012d81c, DcmStack & pixelStack={...})  Line 384 + 0x2f bytes	C++
 	theapp.exe!DcmPixelData::chooseRepresentation(E_TransferSyntax repType=EXS_JPEGProcess1TransferSyntax, const DcmRepresentationParameter * repParam=0x0012d81c, DcmStack & pixelStack={...})  Line 294 + 0x31 bytes	C++
 	theapp.exe!DcmDataset::chooseRepresentation(E_TransferSyntax repType=EXS_JPEGProcess1TransferSyntax, const DcmRepresentationParameter * repParam=0x0012d81c)  Line 528 + 0x2b bytes	C++

[tclune]

I don't know whether this is your problem, but your initialization looks wrong. You set bits stored to 7. It looks like that should be 8, especially if you're going to claim that the max value can be 255. (BTW, high bit should be 7 -- that value starts counting from 0. So don't change that one.) FWIW.

[zaq]

Yes, you're right. Thank you for pointing that out, I must have misread the DICOM. I can't tell you now if that helped with the main issue -- I am in the middle of developing -- but as soon I get the project built I'll check that.

BTW I usually validate datasets against DVT's (Dicom Validation Toolkit) Store SCP and it didn't result even with a warning. It's a good lesson for the future.

[Marco Eichelberg]

Make sure that your application includes

Code: Select all

#include "dcmtk/dcmimage/diregist.h"     /* include to support color images */

That will enable support for color images within the DicomImage code, which is indeed required by the JPEG codec. Also note that chooseRepresentation() will indeed initiate the JPEG encoder (DJCodecEncoder::encode) which will only work correctly if the image pixel module is complete - your sample code does not show that you actually have PixelData in the dataset. The JPEG codec will certainly insist on having pixel data to compress

[zaq]

Make sure that your application includes

Code: Select all

#include "dcmtk/dcmimage/diregist.h"     /* include to support color images */

That is exactly what I needed. Thank you very much for your help, everything works fine now.