Luis Filipe Nassif

Firstly, thank you for this great library! The original issue that originated this is: https://github.com/sepinf-inc/IPED/issues/1203 Analyzing the heap dump, it is possible to see some classes of zxing taking about...

As noticed by @hauck-jvsh, possibly the image background is generating many many qrcode candidate combinations, taking a long time to evaluate and consuming lots of memory resources. Maybe a possible...

> Oh I totally ignored that, it's not a barcode? surely this isn't what you're feeding to the library. But, sure, that's a huge image with no barcode in it....

> Can you share anything from the heap dump that suggests where the allocation takes place? I'll try to generate the heap dump, upload somewhere and post the link here....

@srowen I really appreciate your attention, thank you very much. As an open source developer, I know users/devs usually don't appreciate personal time dedicated to open source projects like this....

Awesome, thank you @srowen for digging into this so fast! For sure, we will test your fix tomorrow with other similar triggering files. We already switched to single QR code...

I couldn't wait :-). The fix worked on other 5 similar triggering files, thanks! I'll run on a large test set through the night to check for regressions on detection.

Hi @srowen, the test on ~500k images finished. From 14 unique QR codes found by version 3.4.1, the 3.5.1-snapshot with the fix missed 2. I'm attaching the images here: [missed-qrcode.zip](https://github.com/zxing/zxing/files/9055953/missed-qrcode.zip)

Thank you @srowen, don't worry, I just reported for a possible future improvement, I think it's better with the fix than without it.

Any plans to integrate this PR? We can not upgrade the library because we need thumbnail extraction too, that method removal was a breaking change without pointing to alternative ways...