Optical character recognition, also known as OCR, is a method for recognizing a word or a phrase in scanned images. It has been developed through years of research. It has had great success in detecting text on scanned images. However, it does not give the desired result in natural images. Therefore, it is necessary to develop special approaches to detect texts in natural images. This study used Otsu and The Maximum Stable Extremal Regions (MSER) image segmentation methods to detect regions with text on natural images. Image segmentation is dividing an image into meaningful regions to analyze it better. In the Otsu model, the most appropriate threshold value is determined for the image, and the image is divided into two classes, foreground, and background, according to this threshold value. On the other hand, the MSER method blocks non-text regions and encloses regions thought to be text in bounding boxes. The study carried out aimed to determine the text areas on 20 natural images selected from the ICDAR 2013 data set with the Otsu method and the MSER method. After segmentation on the natural image, OCR was applied to the images to detect the text on the natural images, and the accuracy rates were compared.
Otsu’s method Maximally Stable Extremal Regions Optical Character Recognition
Otsu modeli Maksimum Kararlı Ekstrem Bölgeler Optik Karakter Tanıma Otsu’s method Maximally Stable Extremal Regions Optical Character Recognition
Birincil Dil | Türkçe |
---|---|
Konular | Mühendislik |
Bölüm | Makaleler |
Yazarlar | |
Yayımlanma Tarihi | 26 Aralık 2022 |
Yayımlandığı Sayı | Yıl 2022 Cilt: 10 Sayı: 5 |