Robust detection of skew in document images.

Area of Science:

• Computer Vision
• Image Processing
• Document Analysis

Background:

• Document skew is a common issue in scanned images.
• Accurate skew detection is crucial for subsequent document processing tasks.
• Existing methods may be slow or require pre-segmentation.

Purpose of the Study:

• To develop a fast and robust algorithm for skew detection in binary document images.
• To improve the efficiency and applicability of skew detection methods.
• To present experimental validation on diverse document types.

Main Methods:

• The algorithm employs interline cross-correlation analysis.
• Correlation is calculated over randomly selected small regions, avoiding full image processing.
• A Monte Carlo sampling technique determines the optimal number of regions for analysis.
• Maximum median of cross-correlation is used as the skew detection criterion.

Main Results:

• The proposed algorithm demonstrates robustness across various document types.
• It effectively detects skew in documents with different linguistic scripts.
• The method achieves high accuracy without prior text/graphics segmentation.
• Experimental results confirm the algorithm's speed and reliability.

Conclusions:

• The developed algorithm offers an efficient and accurate solution for document image skew detection.
• It eliminates the need for complex pre-processing steps like segmentation.
• The method is suitable for a wide range of document images, enhancing automated document analysis workflows.