Text String Detection from Natural Scenes by Structure-based Partition and Grouping

By Chucai Yi, Yingli Tian. IEEE Transactions on Image Processing, 2011

402 citations recorded by the source corpus.

Text information in natural scene images serves as important clues for many image-based applications such as scene understanding, content-based image retrieval, assistive navigation, and automatic geocoding. However, locating text from a complex background with multiple colors is a challenging task. In this paper, we explore a new framework to detect text strings with arbitrary orientations in complex natural scene images. Our proposed framework of text string detection consists of two steps: 1) image partition to find text character candidates based on local gradient features and color uniformity of character components and 2) character candidate grouping to detect text strings based on joint structural features of text characters in each text string such as character size differences, distances between neighboring characters, and character alignment. By assuming that a text string has at least three characters, we propose two algorithms of text string detection: 1) adjacent character grouping method and 2) text line grouping method. The adjacent character grouping method calculates the sibling groups of each character candidate as string segments and then merges the intersecting sibling groups into text string. The text line grouping method performs Hough transform to fit text line among the centroids of text candidates. Each fitted text line describes the orientation of a potential text string. The detected text string is presented by a rectangle region covering all characters whose centroids are cascaded in its text line. To improve efficiency and accuracy, our algorithms are carried out in multi-scales. The proposed methods outperform the state-of-the-art results on the public Robust Reading Dataset, which contains text only in horizontal orientation. Furthermore, the effectiveness of our methods to detect text strings with arbitrary orientations is evaluated on the Oriented Scene Text Dataset collected by ourselves containing text strings in nonhorizontal orientations.

Subfields: Computer Vision and Pattern Recognition, Media Technology

View Paper: https://www.semanticscholar.org/paper/f6b2e7b3ccf75e201bea45c553417798d6dd6610

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

In Media Technology

Citations in the corpus

Detecting Curve Text in the Wild: New Dataset and New Solution (2017, background): “After 2010, the multi-oriented datasets with rotated rectangular labels (NEOCR [23], OSTD [36], MSRA-TD500 [35] and USTB-SV1K [37]) come out, which stimulate many inﬂuential multi-oriented detecting methods in the literatures.”
Symmetry-based text line detection in natural scenes (2015, methodology): “These deep learning based methods [37, 11, 10] usually achieve superior performance over conventional methods [7, 25, 47, 29, 41, 26].”
A Fast Uyghur Text Detector for Complex Background Images (2018, neutral): “veloped [23]–[25].”
Text Detection and Recognition in Imagery: A Survey (2015, methodology): “In [150], Yi and Tian proposed using a Hough Transform to fit text lines to the centroids of components.”
Multi-oriented Text Detection with Fully Convolutional Networks (2016, background): “oriented text detection in the wild is first studied by [31, 29].”
Scene text detection and recognition: recent advances and future trends (2015, methodology): “[28] can detect tilted texts in natural images.”
A robust arbitrary text detection system for natural scene images (2014, methodology): “…robust reading competition data, Street View Text data (SVT) Wang & Belongie, 2010, KAIST scene text data (Lee, Cho, Jung, & Kim, 2010), Microsoft data (Epshtein et al., 2010), Oriented Scene Text Database (OSTD) Yi & Tian, 2011 and MSRA Text Detection 500 data (MSRA-TD500) (Yao et al., 2012).”
Scene Text Detection and Recognition: The Deep Learning Era (2018, methodology): “In this period of time, most text detection methods either adopt Connected Components Analysis (CCA) [24], [52], [58], [98], [135], [156], [159] or Sliding Window (SW) based classiﬁcation [17], [70], [142], [144].”
A Unified Framework for Multioriented Text Detection and Recognition (2014, methodology): “Previous methods either only tackle one of these two tasks [5], [9], [13], [18], [20] or treat them as two isolated stages in an end-to-end text recognition system [6], [17], [21].”
Deep Matching Prior Network: Toward Tighter Multi-oriented Text Detection (2017, methodology): “Recently, numerous techniques [35, 36, 13, 39] have been devised for multi-oriented text detection; these meth-ods used rotated rectangle to localize oriented text.”
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.”
Robust Scene Text Detection with Convolution Neural Network Induced MSER Trees (2014, background): “intensity, stroke width, or color) to construct component candidates [23,24,22,34,7,31,10,32,2].”
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].”
Detecting texts of arbitrary orientations in natural images (2012, methodology): “In [29, 26], methods that can detect text strings of arbitrary directions are proposed but they have a large set of rules and parameters; how general and applicable they are in dealing with scenes of large variation is unclear.”
Real-time scene text localization and recognition (2012, methodology): “The proposed method however significantly outperforms the second best (published) method of Yi [23] in all three measures (see Table 2).”