Morphological background detection and illumination normalization of text image with poor lighting

Area of Science:

• Computer Vision
• Image Processing
• Document Analysis

Background:

• Text images often suffer from uneven illumination due to poor lighting conditions.
• This non-uniform lighting significantly degrades image quality and hinders accurate text recognition.
• Existing methods may not adequately address complex illumination variations.

Purpose of the Study:

• To develop and validate an illumination normalization algorithm for text images with uneven backgrounds.
• To improve the robustness of text detection and recognition systems under adverse lighting.
• To present a novel approach based on morphological transformations for illumination correction.

Main Methods:

• The core method utilizes morphological Top-Hat transform for illumination normalization.
• A three-procedure approach is employed: classical Top-Hat, multi-direction illumination concept using reconstruction, and merging results.
• Opening by reconstruction and closing by reconstruction with multi-direction structuring elements are utilized for optimization.

Main Results:

• The proposed algorithm effectively normalizes uneven illumination in text images.
• Verification through synthetic and real-world images demonstrates successful performance.
• The method produces an even illumination result suitable for subsequent text analysis.

Conclusions:

• Morphological transformations, particularly optimized Top-Hat transform, provide an effective solution for uneven illumination normalization.
• The multi-direction approach enhances the algorithm's ability to handle complex lighting variations.
• This technique significantly improves the quality of text images captured under poor lighting conditions.