[BUG] Not able to scan the barcode of this image using this library, however able to scan it with the phone app about zbar-wasm

[undecaf]

IMHO the quality of that image is just too low for being scanned by zbar.

Could you please tell me which online barcode readers you were using and which barcode type they reported? Thank you.

[kotasudhakar]

These were all able to scan

https://play.google.com/store/apps/details?id=com.gamma.scan&hl=en&gl=US
https://play.google.com/store/apps/details?id=com.mobileappsshop.barcode
https://online-barcode-reader.inliteresearch.com/
https://nanonets.com/barcode-scanner

[undecaf]

Thank you @kotasudhakar for the list.

https://online-barcode-reader.inliteresearch.com/ reports a module size of 1.8px which makes this quite an edge case.

https://nanonets.com/barcode-scanner did not scan the barcode at all, it just detected the characters above and below the barcode.

Since zbar-wasm uses the Zbar barcode reader (compiled to WASM) in the background, there is nothing I can do in order to improve detection quality. I can only suggest that you open an issue at that project that suggests to improve detection for barcodes with very small module sizes.

[zcattacz]

Hi @undecaf, maybe have a try on this one https://zxing-wasm-demo.deno.dev/, when binarizer is switched to FixedThrethold mode, it works. Clicking on the green circle on the upper right corner displays the decoded detail.

[undecaf]

Thank you @kotasudhakar, @zcattacz for your contributions to this issue.

IMHO, it is not a bug if a scanner is not able to recognize a barcode because of low image quality -- it is just a consequence of the amount of detection effort in the scanning algorithm.

zbar-wasm is intended to be fast enough to scan barcodes in mobile browsers from the video stream of the camera. If real-time scanning is not required, detection quality could be improved by preprocessing images in JS before passing them to the scanner, but this is outside the scope of zbar-wasm.

Therefore I hope you understand that I am closing this issue.