Innovative Edge Detection Approaches for Arabic Character Recognition in Kitab Kuning Images: A Comparison of Sobel, Prewitt, and Canny

Abstract

The preservation of Kitab Kuning manuscripts as part of Islamic intellectual heritage requires digitalization and image enhancement to ensure readability and sustainability. One of the essential stages in digital manuscript processing is edge detection, which plays an important role in extracting the structural shape of Arabic characters. This study aims to compare the performance of three classical edge detection methods—Sobel, Prewitt, and Canny—in detecting the edges of Arabic characters in Kitab Kuning manuscript images. Each manuscript image was preprocessed through grayscale conversion and Gaussian filtering before applying the three edge detection algorithms. The resulting edge images were evaluated using Mean Square Error (MSE), Peak Signal to Noise Ratio (PSNR), and Structural Similarity Index (SSIM). The experimental results show that the Canny method consistently produces lower MSE values, higher PSNR, and more stable SSIM compared to Sobel and Prewitt, resulting in clearer, smoother, and more accurate edge structures resembling the original Arabic characters. Prewitt demonstrates moderate performance, while Sobel tends to generate rougher edges and is more sensitive to noise. Based on these findings, the Canny method is recommended as the most effective approach for Arabic character edge detection in the digital preservation of Kitab Kuning manuscripts, supporting better quality enhancement and long-term usability of digital Islamic manuscripts