Define Custom Recognition Blocks on Document Image & Improved OCR Recognition
Join the DZone community and get the full member experience.Join For Free
What’s new in this release?
We are pleased to announce the release of Aspose.OCR for Java 2.0.0. This release contains 12 improvements, including the showcased feature discussed below. Setting custom recognition blocks is useful when documents of the same layout have to be processed. Programmers can specify a custom recognition blocks and direct the API to focus on a particular area of the image. The latest build allows the users to define custom recognition blocks on the document image. These blocks are processed by the OcrEngine while other areas are skipped completely. If user-defined recognition blocks are not set prior to processing, the OcrEngine automatically calculate the recognition blocks using its default behavior. Images with text can be divided into many text blocks on which recognition process can be performed. It is sometimes important to apply a special sorting so that the correct text block order is maintained. The OcrConfig class has exposed the DetectReadingOrder property to handle situations where an image within the main document image may contain text blocks of its own. Setting the DetectReadingOrder property to true instructs the OcrEngine to maintain the order of text regions for better recognition result. However, performance degradation can be observed for images with a large number of text blocks. In such cases, it is advised to turn this feature off by setting the DetectReadingOrder property to false. In order to improve the accuracy of the recognized data, the Aspose.OCR for Java API has exposed the DoSpellingCorrection property for the OcrConfig class. Setting the DoSpellingCorrection property to true will improve the OCR results, however, the process may take more time to complete. Please check the detailed article on setting the OcrEngine to automatically correct spelling. Document images to be processed with OcrEngine may contain graphics. During the OCR process, these graphics may not contain any text to be recognized although they are detected as a separate block. If developers wish to ignore the non-textual blocks such as graphics from the recognition process, they just have to set the RemoveNonText property exposed by the OcrConfig class to true. Last but not the least; we have completely overhauled the OcrEngine with new recognition algorithm for better accuracy of the recognized data and performance considerations. Moreover, Aspose.OCR API now uses a resource file that is just of 5.5MB in size unlike previous resource files of 88MB. This release includes plenty of new & enhanced features as listed below
- New OCR algorithm is an overhaul of the OCR engine used in the component. It provides better recognition quality and better performance overall.
- Support detection of reading order.
- Support automatic text regions detection.
- Support spelling correction.
- Allow user to define which blocks to process.
- Support removing of non-text blocks.
- Add possibility to recognize text on complex background
- Resource file path should be specified explicitly by user.
- Restore layout of image
- [Java] Aspose.OCR & Aspose.OMR throws AsposeLicenseException while setting a valid license
- [Java] Fully Test Aspose.OCR for Java with Java 8
Newly added documentation pages and articles
Some new tips and articles have now been added into Aspose.OCR for Java documentation that may guide you briefly how to use Aspose.OCR for performing different tasks like the followings.
Overview: Aspose.OCR for Java
Aspose.OCR for Java is a character recognition component that allows developers to add OCR functionality in their Java web applications, web services and Windows applications. It provides a simple set of classes for controlling character recognition tasks. It helps developers to work with image files from within their Java applications. It allows developers to extract text from images, Read font, style information quickly, saving time & effort involved in developing an OCR solution from scratch.
Opinions expressed by DZone contributors are their own.