Enhancing Text Recognition Performance Through Multi-Dimensional Data Analysis

• Published in: 2024 5th International Conference on Image Processing and Capsule Networks (ICIPCN) pp. 787 - 792
• Main Author: Zhu, Huiming
• Format: Conference Proceeding
• Language: English
• Published: IEEE 03.07.2024
• Subjects: Accuracy; Algorithm optimization; Analytical models; Data analysis; Image recognition; Machine learning; Multi-dimensional data mining; Refining; Text detection; Text mining; Text recognition
• DOI: 10.1109/ICIPCN63822.2024.00136

This study presents a novel methodology for improving text recognition accuracy through multidimensional data analysis. By examining state-of-the-art algorithms and methods, the technical and practical challenges associated with text detection and recognition in natural scenes are identified and addressed. The approach presents innovative text region detection algorithms and text recognition models, complemented by differentiable binarization techniques and multidimensional data analysis. Extensive simulations and experiments validate the effectiveness of the approach, demonstrating significant improvements in accuracy over traditional methods across a wide range of datasets and scenarios. Specifically, experiments show that the proposed multidimensional data analysis approach achieves an average accuracy of about 98.35%, surpassing the average accuracy of about 95.99% achieved by traditional one-dimensional models. Furthermore, when comparing different methods in text region detection, the proposed approach consistently outperforms existing methods, with an average accuracy of about 98.77%, compared to about 95.85% and 96.19% for the alternative methods, respectively. These results underscore the superiority of the methodology in addressing the challenges of text recognition in natural scenes.