16.2.6+103
📅 2025-05-16
Improvements
Text direction hint for OCR
The possibility of specifying a text direction for OCR of Asian languages has been reintroduced to improve recognition accuracy in specific scenarios.
As a consequence, the value TextDirection.Any has also been added to the relevant enum to represent automatic detection, which remains the default behavior.
Default vs. Explicit behavior
-
Default behavior: The OCR engine automatically detects the text direction (
TextDirection.Any). This will give the appropriate output in the vast majority of the use cases. -
Explicit behavior: In certain scenarios — especially in the context of zonal OCR — automatic detection may misinterpret the direction due to limited context. In these cases, explicitly setting the text direction when already known will improve accuracy.
The following snippet demonstrates how to configure OCR with an explicit text direction:
COcrContext objContext = COcrContext::Create(evLanguage);
objContext.SetTextDirection(TextDirection::TopToBottom);
CTextRecognition objTextRecognition = CTextRecognition::Create(objIdrs);
objTextRecognition.SetOcrParams(COcrPageParams::Create(objContext));
objTextRecognition.RecognizeText(objImage);Greyscale detection during PDF loading
A new property, CImageLoadOptionsPdf.GreyscaleDetection, has been introduced in order to enable/disable the rasterization of PDF pages as greyscale CImage objects.
This property is enabled by default, maintaining behavior consistent with previous iDRS releases.
Disabling greyscale detection can offer performance improvements in terms of speed, but at the cost of increased memory usage, as pages will be rasterized in color rather than greyscale.
Fixed bugs
| Internal ID | Description | Service desk IDs |
|---|---|---|
IDRSRD-9934 |
PDF/UA generated by the iDRS doesn’t succeed compliance validation when document contains complex tables |
|
IDRSRD-9927 |
the iDRS encounters a crash when rotating a specific image |
|
IDRSRD-9920 |
The iDRS requires extra unexpected OCR resources to run auto-orientation only |
|
IDRSRD-9908 |
The iDRS should expose a flag to enable or disable greyscale image detection during loading of a PDF page |
|
IDRSRD-9903 |
Confidence values of language and orientation detection feature are unusable with 16.2.6+82 |
ISD-36788 |
IDRSRD-9901 |
The default character set is missing some supported characters for Japanese language |
|
IDRSRD-9900 |
.NET samples cannot be compiled on Linux |
|
IDRSRD-9884 |
Graphic shapes detected by iDRS are incorrectly scaled on output document if input image resolution is different than 300 dpi |
|
IDRSRD-9883 |
The OCR engine library is linked with WS2_32.dll for no reason |
ISD-36751 |
IDRSRD-9882 |
API reference main pages are different between C++, .NET and C APIs |
|
IDRSRD-9856 |
The iDRS can return OCR results outside of input zones when running zonal OCR |
|
IDRSRD-9839 |
The iDRS merges lines from different text columns on a specific image |
|
IDRSRD-9796 |
Orientation detection gives unexpected answer on a border-case scenario |
|
IDRSRD-9795 |
An integrator should be able to hint the OCR engine for the text direction to detect, when processing Asian documents |
|
IDRSRD-9754 |
The iDRS is not compatible with VirtualBox VMs running on Windows Hosts |
ISD-36479 |
IDRSRD-9703 |
The new segmentation filters isolated punctuations or characters |
|
IDRSRD-5619 |
DOCX output created by the iDRS is poor for specific images |