A Clustering-Based Method for Detecting Text Area in Videos Recorded with the Aid of a Smartphone

Abstract

Text detection is a crucial task in image processing and computer vision applications. Several methods have been proposed, but little attention has been given to the detection of text area in the video frame recorded with the aid of a smartphone. To answer this gap, we propose in this work a new method for text detection in video acquired by a smartphone. The method is based on the use of the Line Segment Detector (LSD) algorithm and the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) algorithm to detect the line segments that belong to the text region. Subsequently, we carry out a set of filters to select the four segments that represent the four sides of the text area in each video frame. The experimental results on ICDAR2015 Smartphone Document Capture Dataset demonstrate that the proposed method provides better performance and achieves promising detection accuracy.