TEXT DETECTION FROM IMAGES USING MATLAB-BASED IMAGE PROCESSING
Abstract
Text localization in digital images, using MATLAB simulation. The system aims to bridge the gap between the low level (pixel) and high level (semantic) by identifying regions of interest (ROIs) in which text can be found. The system uses a suite of image processing algorithms starting with the detection of high frequency intensity gradients using a Sobel edge detector. In order to overcome the problem of discontinuous edges of individual characters, the system uses Mathematical Morphology, in particular, dilatation with a rectangular structuring element (3 \times 20) to join disjoint features and create text blocks. These are then extracted using automatic hole removal and area filter to discard any background noise. The main innovation of this simulation is the Heuristic Evaluation Model which measures the proposed regions by geometric constraints. A weighted probability is given to each candidate block as its Solidity (compactness) and Extent (rectangularity). The experiment with several trials demonstrates the system's high localization accuracy, and provides an average accuracy of 70.67% for approximate matches. This work has shown how morphological operations, as traditionally used, can provide data to subsequent Optical Character Recognition (OCR) systems that are appropriate and save time in complex scenes.
References
Gonzalez, R. C., & Woods, R. E. (2018). Digital Image Processing (4th ed.). Pearson.
Jain, A. K. (1989). Fundamentals of Digital Image Processing. Prentice-Hall, Inc.
Shi, Z., Setlur, S., & Govindaraju, V. (2013). "A Survey of Text Detection and Recognition in Images and Video." IAPR International Conference on Document Analysis and Recognition.
Keshavarz Ghorabaee, M., et al. (2015). "EDAS: A novel multi-criteria decision-making approach for supplier evaluation and selection." Economic Research-Ekonomska Istraživanja.
Ye, Q., & Doermann, D. (2015). "Text Detection and Recognition in Imagery: A Survey." IEEE Transactions on Pattern Analysis and Machine Intelligence.
MathWorks Documentation (2024). "Detect and Recognize Text in Images." Computer Vision Toolbox User’s Guide. (Technical reference for the regionprops and imfill functions used in your code).
Cheriet, M., et al. (2007). Character Recognition Systems: A Guide for Students and Practitioners. Wiley-Interscience. (Background on why Solidity and Extent are vital for text vs. noise separation).
Soille, P. (2013). Morphological Image Analysis: Principles and Applications. Springer Science & Business Media. (Detailed explanation of how dilatation and erosion affect object segmentation).
Canny, J. (1986). "A Computational Approach to Edge Detection." IEEE Transactions on Pattern Analysis and Machine Intelligence. (Provides a comparative background to the Sobel method used in your simulation).
Long, S., He, X., & Yao, C. (2021). "Scene Text Detection and Recognition: The Deep Learning Era." International Journal of Computer Vision. (Useful for the 'Future Scope' or 'Literature Review' to contrast your classical method with modern AI approaches).
Satzoda, R. K., & Trivedi, M. M. (2014). "On Performance Evaluation Metrics for Object Detection." IEEE Conference on Computer Vision and Pattern Recognition Workshops. (Provides a basis for using confidence scores in object localization).
Szeliski, R. (2022). Computer Vision: Algorithms and Applications. Springer Nature. (The primary textbook for understanding the mathematical foundations of the Sobel operator and image gradients).
