10.2

Edit

Issues related to OCR results quality

If you noticed that the quality of the current OCR results is unsatisfying or degraded as compared to previous ones, act as follows to investigate the issue:

Analyze task processing parameters

Check the parameters used for task processing. For that, look up the respective table in the database.

note

For the check, you need access to the MS SQL database.

The sequence of steps for the check is as follows:

  1. Get the taskId for your document. You can find it in the ocr_task_id column in the results table after OCR steps in Control Tower.
  2. Once you get the taskId, connect to the MS SQL database and choose the workfusion database.
  3. Switch to the OCR schema and look at the abbyy_task table, particularly the task_parameters column that should contain all used parameters.
  4. Compare the parameters with previous runs.

Also, pay attention to used FreEngine. It can have version 11 (v1) or version 12 (v2). For details, refer to Switch ABBYY OCR API from FRE11 to FRE12.

Evaluate input document quality

Check the original document quality, paying special attention to the following characteristics and parameters:

  • Resolution: if it's too low, you can manually adjust it using the changeDPI parameter.
  • Colored background or document parts: it's strongly recommended to use black-and-white documents to achieve the best results of OCR processing. Pay attention to the removeColorObjectsType parameter.
  • Garbage: use the removeGarbageSize parameter for scanned images to remove garbage—excessive dots smaller than a specific size.
  • Contrast: check the enhanceLocalContrast parameter.
  • Content types: analyze your document to identify whether it contains tables, pictures, barcodes, text, and so on. Consider using the allowedRegionTypes parameter.

For more details, refer to the Improve OCR results article.

To achieve the most accurate and detailed comparison and results, try testing the OCR process as described below:

  • On the same document (the most critical option)
  • With the same parameters

For OCR troubleshooting instructions, refer to the guide.

Read also:

← Issues related to unsupported fonts, languages, formatIssues related to no OCR results →