Our algorithm relies on Document-oriented Local Features (DoLF) [Zagoris2017, Zagoris2014], which considers information around representative keypoints and a matching process that incorporates spatial context in a local proximity search without using any training data. Finally, it introduces a distance algorithm that incorporates spatial context and is employed under segmentation-based and segmentation-free scenarios.
The main novelties of the above approach are:
- Use of local features that take into consideration the handwritten document particularities. Therefore, it can detect the characters' meaningful points that reside in the documents independently of its scaling.
- It provides consistency between different handwritten writing variations.
- Use of the same operational pipeline in both segmentation-based and segmentation-free scenarios
- Incorporation of spatial context in the local search of the matching process.
The segmentation - free operational pipeline:
 |
 |
|---|---|
| the query image, the localpoints, the central location (shown in magenta color), and its nearest keypoint (shown in orange color) | the document image |
 |
 |
| the candidate local points for the document coordinate origin | multiple instances of word boundaries around each candidate coordinate origin |
 |
 |
| multiple word detection | final result (the green color denotes the most similar word) |
Finally, implementing the proposed keyword spotting method as a recommender system to a transcription process is available at http://vc.ee.duth.gr/ws [Zagoris2015].
A more efficient matching procedure that uses the DoLF local points is available at http://orpheus.ee.duth.gr/word-spotting-demonstrator/
WordSpottingByExample 1.0.0.0
indexing Indexing Directory
retrieval Retrieve Word
help Display more information on a specific command.
version Display version information.WordSpottingByExample 1.0.0.0
USAGE:
Indexing example:
WordSpottingByExample indexing --imageformat jpg C:\WordImages .\dataset.sqlite
-i, --imageformat (Default: png) Image Extension
--help Display this help screen.
--version Display version information.
ImagesDirectory (pos. 0) Required. The directory path that contains the document images for indexing
Output SQLite Dataset File (pos. 1) Required. The output SQLite file that contains the dataset info
WordSpottingByExample 1.0.0.0
USAGE:
Locate a word:
WordSpottingByExample retrieval .\query-word.png .\dataset.sqlite Results.xml
-i, --imageformat (Default: png) Image Extension
--help Display this help screen.
--version Display version information.
QueryImagePath (pos. 0) Required. The query word image path. It must be a directory or a file
Dataset (pos. 1) Required. The SQLite database file
Results (pos. 2) Required. The XML Retrieval Results File. It follows the H-KWS2014 XML Format. Download
Evaluation Tool from <https://vc.ee.duth.gr/H-KWS2014/#VCGEval>
[Zagoris2014] K. Zagoris, I. Pratikakis and B. Gatos, "Segmentation-Based Historical Hand-written Word Spotting Using Document-Specific Local Features," 2014 14th International Conference on Frontiers in Handwriting Recognition, Heraklion, 2014, pp. 9-14.
[Zagoris2015] K. Zagoris, I. Pratikakis, and B. Gatos, "A framework for efficient transcription of historical documents using keyword spotting," in Historical Document Imaging and Processing (HIP'15), 3rd International Workshop on, August 2015, pp. 9–14.
[Zagoris2017] K. Zagoris, I. Pratikakis, B. Gatos. 2017 Unsupervised Word Spotting in Historical Handwritten Document Images using Document-oriented Local Features. Transactions on Image Processing. Under Review.