IMPLEMENTASI MSER DAN OPTICAL CHARACTER RECOGNITION (OCR) UNTUK DETEKSI TEKS PADA GAMBAR

Abstract

This research develops a text detection system in images by implementing the integration of Maximally Stable Extremal Regions (MSER) and Optical Character Recognition (OCR) methods. The main problem addressed is the limitation of text detection accuracy in images with complex background variations, different resolutions, and uneven lighting conditions. The research methodology involves data collection of 100 image samples from signboards in the surrounding environment, followed by preprocessing stages including MSER implementation for identifying stable regions with similar pixel intensity that potentially contain text, and OCR application for recognizing text from extracted regions. The system testing was conducted using confusion matrix with precision, recall, and accuracy parameters. The research results show that the developed system successfully achieved high performance with precision of 98%, recall of 94%, and accuracy of 94%. The MSER method proved effective in detecting text candidate regions despite variations in font, size, and orientation, while OCR demonstrated good capability in character recognition from the detected regions. This integration provides a robust and practical solution for automatic text detection applications in various real-world scenarios.