Fixed bug in parsing of JPEG DICOM data in MINT ref. impl.

[Uli Bubenheimer]

A few weeks ago I fixed a bug in the MINT reference implementation related to parsing of DICOM instances that are in encapsulated JPEG format. In some cases the parsing of the multi-frame images did not split them up correctly. I have not seen a failure in an actual dataset, I found this problem through code analysis.

The fix is on trunk in revision 711 (Dec. 6 2011).

Here is the scenario where I believe this would have caused a problem; images may get truncated, may not be displayable, or may be missing:

1. The DICOM instance must have one of the JPEG transfer syntaxes.
2. The Basic Offset Table of the Pixel Data tag must be empty.
3. There must be a frame f which is encoded in more than one fragment.
4. There must be additional frames after f.
5. The value of one of the fragments of f other than the first fragment must start with byte value 0xD8 or 0x4F.
6. Either the fragment in question is not the last fragment of f, or at least one of the frames after f has multiple fragments.

Uli