Publication: Enhancing image-to-text recognition using image processing techniques
No Thumbnail Available
Date
2023-05
Journal Title
Journal ISSN
Volume Title
Publisher
Don Mariano Marcos Memorial State University - South La Union Campus
Abstract
Image recognition has gained considerable attention in the field of research. If
attracted researchers to experiment on its application, impact, and contribution to different
scientific fields. Several applications with image recognition are available nowadays.
However, their performance results still needed improvement especially in terms of
detecting and recognizing text images from signages or scene texts with blurry images.
With this, the researchers aimed to enhance the image-to-text recognition of the existing
algorithm using image pre-processing techniques such as blind deconvolution, sharpening,
or a combination of both. Specifically, the researchers wanted to compare the performance
of image-to-text-recognition of the existing algorithms before and after applying preprocessing steps, Improve the existing algorithm s performance, and compare the existing
algorithm's performance to the current state-of-the-art Scene Text Recognition (STR) by
computing its Character Recognition (CER), Word Error Recognition (WER), Confidence
Score and Minimum Edit Distance (MED). The proposed techniques showed a significant
improvement in text recognition by enhancing the quality of the images. This indicated that
the sharpening process has successfully enhanced the clarity and sharpness of the text,
resulting in a more accurate recognition. The proposed wiener filtering achieved a 0.35%
higher than the baseline and the proposed unsharp masking 0.16% higher than the
baseline.
Description
Full text.
Keywords
Citation
Orencia, J. C., Hipol, C. J., Madriaga, J. C., Mosico, M. T. A. G., & Ubungen, K. M. P. (2023) Enhancing image-to-text recognition using image processing techniques [Unpublished Undergraduate thesis]. Don Mariano Marcos Memorial State University - South La Union Campus, Agoo, La Union.