Robust text detection in natural images with edge-enhanced Maximally Stable Extremal Regions

By Huizhong Chen, Sam S. Tsai, Georg Schroth, David M. Chen, Radek Grzeszczuk, Bernd Girod. 2011 18th IEEE International Conference on Image Processing, 2011

494 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Media Technology

View Paper: https://www.semanticscholar.org/paper/6cb3153e5773053916a27bf3ab4530705a6bcf80

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

1. Xiang Bai
2. Andrew Zisserman
3. Cong Yao
4. Yu Qiao
5. Luc Van Gool
6. Andrea Vedaldi
7. Chunhua Shen
8. Dacheng Tao

In Media Technology

1. Cong Yao
2. Xiang Bai
3. Yu Qiao
4. Andrea Vedaldi
5. Minghui Liao
6. Andrew Zisserman
7. Tong He
8. Weilin Huang

Citations in the corpus

• End-to-End Text Recognition with Hybrid HMM Maxout Models (2013, methodology): “The text recognition problem has been addressed in the literature on multiple levels: character recognition (Saidane & Garcia, 2007; Coates et al., 2011), word recognition (Novikova et al., 2012; Mishra et al., 2012) and text detection (Hanif et al., 2008; Chen et al., 2011).”
• TextProposals: A text-specific selective search algorithm for word spotting in the wild (2016, methodology): “This way, the hierarchical structure of text has been traditionally exploited in a post-processing stage with heuristic rules [18, 26, 33, 25], usually constrained to search for horizontally aligned text in order to avoid the combinatorial explosion of enumerating all possible text lines.”
• Scene text detection and recognition: recent advances and future trends (2015, methodology): “SWT [9] and MSER [10] are two representative methods in the field of scene text detection, which constitute the basis of a lot of subsequent works [12–14, 29, 30, 34, 48, 49].”
• Residual Enhanced Visual Vectors for on-device image matching (2011, background): “Many mobile visual search (MVS) applications have been successfully developed for recognition of outdoor landmarks [1], product covers [2], and printed documents [3], [4], amongst other categories.”
• Deep Direct Regression for Multi-oriented Scene Text Detection (2017, background): “Most scene text detection methods [26] [21] [8] [1] [17] treat text as a composite of characters, so they first localize character or components candidates and then group them into a word or text line.”
• A Unified Framework for Multioriented Text Detection and Recognition (2014, methodology): “These methods have achieved excellent performance on natural images and inspired numerous researchers in this field [42], [46], [47].”
• Deep Features for Text Spotting (2014, methodology): “Text detection methods are either based on connected components (CCs) [11, 16, 50, 49, 32–35] or sliding windows [40, 7, 39, 45].”
• Text Detection and Recognition in Imagery: A Survey (2015, background): “improve the robustness on various text categories, hybrid features have been applied in text localization [1], [24], [90], [115], [73], [138], [122], [174], [171].”
• Cascaded Segmentation-Detection Networks for Text-Based Traffic Sign Detection (2018, methodology): “There are a lot of methods [36]–[40] inspired by these two methods.”
• Reading Text in the Wild with Convolutional Neural Networks (2014, methodology): “These focus solely on text detection (Anthimopoulos et al. 2013; Chen et al. 2011; Epshtein et al. 2010; Gomez and Karatzas 2013, 2014; Huang et al. 2014; Yi and Tian 2011; Yin et al. 2013), text recognition (Almazán et al.”
• Text-Attentional Convolutional Neural Network for Scene Text Detection (2015, methodology): “The connected component methods have achieved great success in text detection and localization [5], [1], [21], [22], [23], [24], [6], [7], [25], [26], [27], [28], [29], [30].”
• Robust Scene Text Detection with Convolution Neural Network Induced MSER Trees (2014, background): “As pointed out in the literature, most connected component based methods suffer from inherent limitations of low-level filters, which easily cause error connections between multiple characters or with background components in some difficult cases, such as low quality or seriously affected texts [24,”