Wicaksana, Ardi Anugerah (2025) Comparative analysis of otsu method for braille image segmentation / Ardi Anugerah Wicaksana</p>. Diploma thesis, Universitas Negeri Malang.
Full text not available from this repository.Abstract
p Noise is one of the primary challenges in the segmentation of Braille images particularly when dealing with scanned outputs that exhibit low contrast uneven illumination and various visual disturbances. These image conditions hinder the accurate extraction of Braille dots which is essential for converting tactile information into machine-readable text. Traditional image segmentation approaches often fail to cope effectively with such distortions especially in low-resource environments where deep learning models are impractical due to computational limitations. This study aims to address the issue of noise by developing a pixel-level threshold-based segmentation model that can be implemented efficiently in lightweight environments (low-resource implementation). The model is developed using Python and Google Colab and incorporates four classical segmentation techniques Otsu Otsu Inverse Otsu with Morphological Operations and Otsu Inverse with Morphological Operations. Each method was tested on a Braille image dataset and evaluated using six quantitative image quality metrics Peak Signal-to-Noise Ratio (PSNR) Mean Squared Error (MSE) Mean Absolute Error (MAE) Structural Similarity Index (SSIM) Feature Similarity Index (FSIM) and Edge Similarity Index (ESSIM). The results demonstrate that the Otsu method combined with morphological operations achieved the highest PSNR and SSIM scores with values of 27.67 dB and 0.5548 respectively. These values indicate improved image fidelity and structural preservation during the segmentation of Braille dots. In conclusion this classical thresholding approach enhanced with morphological operations proves effective in improving segmentation accuracy and is highly promising for use in Braille recognition systems under resource-constrained environments offering an efficient lightweight and reliable alternative to non-deep learning-based methods. /p
| Item Type: | Thesis (Diploma) |
|---|---|
| Divisions: | Fakultas Teknik (FT) > Departemen Teknik Elektro (TE) > S1 Teknik Informatika |
| Depositing User: | library UM |
| Date Deposited: | 28 Jul 2025 04:29 |
| Last Modified: | 09 Sep 2025 03:00 |
| URI: | http://repository.um.ac.id/id/eprint/400166 |
Actions (login required)
 |
View Item |