Skip to content

jpeglossless bug fixes - SOF3 P in [9, 15] and DHT before SOF3#51

Open
bluna301 wants to merge 1 commit into
NVIDIA:mainfrom
bluna301:bl/jpeglossless_DHT_bugfix
Open

jpeglossless bug fixes - SOF3 P in [9, 15] and DHT before SOF3#51
bluna301 wants to merge 1 commit into
NVIDIA:mainfrom
bluna301:bl/jpeglossless_DHT_bugfix

Conversation

@bluna301

@bluna301 bluna301 commented May 6, 2026

Copy link
Copy Markdown

Addresses JPEGLossless Bugs: SOF3 P in [9,15] and DHT-before-SOF3 Streams.

Tests and anonymized DICOM-derived test files added; all tests passing:

[----------] Global test environment tear-down
[==========] 2253 tests from 146 test suites ran. (89574 ms total)
[  PASSED  ] 2253 tests.

Signed-off-by: bluna301 <luna.bryanr@gmail.com>
@hoang-dt

hoang-dt commented May 8, 2026

Copy link
Copy Markdown

Thank you for submitting this PR. This issue was actually fixed in nvJPEG 13.0.2. If you update nvJPEG (through Python or the CUDAToolkit depending on what is used), these images will be decoded correctly. For nvImageCodec, we will add a check for older versions of nvJPEG to correctly reject these images instead of silently failing.

@bluna301

bluna301 commented Jun 8, 2026

Copy link
Copy Markdown
Author

@hoang-dt - sorry for the late reply, thank you for the review and for communicating the fix. We have been using CUDA 12.9 in development and production, so therefore have been using other CUDA 12 compatible dependencies which is why we were seeing pre-patch behavior.

Adding a guard in nvImageCodec for older nvJPEG versions sounds great to me. Let me know if you need any assistance with applying the fix, happy to help if needed.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants