Multi-Oriented and Multi-Lingual Scene Text Detection With Direct Regression

• Published in: IEEE transactions on image processing Vol. 27; no. 11; pp. 5406 - 5419
• Main Authors: He, Wenhao, Zhang, Xu-Yao, Yin, Fei, Liu, Cheng-Lin
• Format: Journal Article
• Language: English
• Published: United States IEEE 01.11.2018; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Artificial neural networks; Boundaries; Computer vision; Detectors; Feature extraction; Fully convolutional network; Heuristic methods; Manuals; multi-oriented; multi-task; Neural networks; Object detection; Object recognition; Pixels; Post-processing; Proposals; Regression analysis; scene text detection; Stress; Task analysis
• ISSN: 1057-7149; 1941-0042
• DOI: 10.1109/TIP.2018.2855399

Multi-oriented and multi-lingual scene text detection plays an important role in computer vision area and is challenging due to the wide variety of text and background. In this paper, first, we point out the two key tasks when extending convolutional neural network (CNN)-based object detection frameworks to scene text detection. The first task is to localize the text region by a downsampled segmentation-based module, and the second task is to regress the boundaries of text region determined by the first task. Second, we propose a scene text detection framework based on fully convolutional network with a bi-task prediction module, in which one is a pixel-wise classification between the text and non-text and the other is pixel-wise regression to determine the vertex coordinates of quadrilateral text boundaries. Post-processing for word-level detection is based on non-maximum suppression, and for the line-level detection, we design a heuristic line segments grouping method to localize long text lines. We evaluated the proposed framework on various benchmarks, including multi-oriented and multi-lingual scene text data sets, and achieved the state-of-the-art performance on most of them. We also provide abundant ablation experiments to analyze several key factors in building high performance CNN-based scene text detection systems.