DicomImage::processNextFrames() causes memory leak

[takeos]

When a DicomImage instance is constructed for a MONOCHROME2 image, DiOverlay instance is also created as below.

Code: Select all

DiOverlay::DiOverlay(const DiDocument * docu=0x03d41748, const unsigned short alloc=8, const unsigned short stored=8, const unsigned short high=7)  line 70
DiMonoImage::Init(DiMonoModality * modality=0x03d3f6e8)  line 669 + 0x4a bytes
DiMonoImage::DiMonoImage(const DiDocument * docu=0x03d41748, EI_Status status=EIS_Normal)  line 103
DiMono2Image::DiMono2Image(const DiDocument * docu=0x03d41748, EI_Status status=EIS_Normal)  line 48 + 0x3d bytes
DicomImage::Init()  line 241 + 0x3c bytes
DicomImage::DicomImage(DcmObject * object=0x03d414d8, E_TransferSyntax xfer=EXS_LittleEndianExplicit, const unsigned long flags=1025, const unsigned long fstart=0, const unsigned long fcount=1)  line 106

Subsequent processNextFrames() call creates another DiOverlay instance as below, but the first one is not deleted since removeReference() is not called for Overlays[0] in DiMonoImage::Init() before assigning a new pointer to it.

Code: Select all

DiOverlay::DiOverlay(const DiDocument * docu=0x03d41748, const unsigned short alloc=8, const unsigned short stored=8, const unsigned short high=7)  line 70
DiMonoImage::Init(DiMonoModality * modality=0x03d3f6e8)  line 669 + 0x4a bytes
DiMonoImage::processNextFrames(const unsigned long fcount=1)  line 657
DicomImage::processNextFrames(const unsigned long fcount=1)  line 184 + 0x22 bytes

[Michael Onken]

Hi,

thanks for the report. I'm not sure if just deleting and then creating a new one is the best way to solve that issue in DiMonoImage::Init(), like this:

Code: Select all

   if (modality != NULL)
    {
        // cleanup old data for builtin overlays (may be existing from previous frame processed)
        if (Overlays[0] != NULL)
            delete Overlays[0];
        Overlays[0] = new DiOverlay(Document, BitsAllocated, BitsStored, HighBit);

because I'm not sure what should/must be re-used etc. So I will wait for Jörg being the author to fix it (currently not in office). We will post here when we fixed that in the source code repository.

Best regards,
Michael

[Jörg Riesmeier]

Thanks again for the report. Should be fixed now.