A Combined Edge Detection Analysis and Clustering based Approach for Real Time Text Detection

Recently, scene text detection has become an active research topic in computer vision and document analysis because of its great importance. However, due to the number of factors, such as variety in font size, style, colors, language, as well as background noise, blur, and occlusion, makes text detection is still challenging. The proposed methods directly run on the frame image of the recording video to detect text. Sobel operator used to perform edge detection on the image while K-means clustering used to extract text from the background of the image. After the image has been segmented based on the cluster, the text will be identified using the Maximum Stable External Region (MSER). Applications run on Android devices and built by using OpenCV for image retrieval processes, while algorithms written using C++. The experimental result shows 71.79%, 62.11%, and 64.56% for average recall, precision, and F1-score, respectively.