Unsupervised Text Segmentation Using Color and Wavelet Features

Gllavata, Julinda; Ewerth, Ralph; Stefi, Teuta; Freisleben, Bernd

doi:10.1007/978-3-540-27814-6_28

Cited by 19 publications

(18 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It has been demonstrated in [9] that the segmentation and the subsequent steps perform better on a higher resolution than on the original video frame resolution of 72 dpi. Furthermore, the segmentation algorithm performs better, if the text is not too small.…”

Section: Resolution Enhancementmentioning

confidence: 97%

“…The text image which gives the best recognition performance is considered as the output of the system. Gllavata et al [9] regard the text segmentation problem as a clustering process. As in [12], the possible text and background color is defined.…”

Section: Related Workmentioning

confidence: 99%

“…Before proceeding further, all feature components are scaled to the range [0, 1]. Additional details about the feature extraction process can be found in [9].…”

Section: Feature Extractionmentioning

confidence: 99%

“…The idea of segmenting the localized text from complex background employing unsupervised learning methods was already presented in our previous work [9]. In this paper, we further explore and extend this idea by introducing the following new aspects: I) a flexible number of clusters, which depends on the image colors; II) unsupervised fuzzy classification of image pixels using the number of clusters defined in the first step; III) evaluation of the compactness of the clusters and their iterative improvement; IV) determination of the cluster where the text pixels are classified and its binarization.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Adaptive Fuzzy Text Segmentation in Images with Complex Backgrounds Using Color and Texture

Gllavata

Freisleben

2005

Lecture Notes in Computer Science

Self Cite

View full text Add to dashboard Cite

Abstract. Textual information present in images can help to achieve the aim of automatic content based annotation and retrieval of images. In this paper, we address the problem of text segmentation (TS) in images with complex background for recognition purposes. The proposed TS method takes as input the localized text and proceeds as follows: First, the number of initial clusters is determined by analyzing the colors of the image. Second, the image pixels are clustered using the number of clusters defined in the first step. The compactness of the clusters is evaluated in each step and improved iteratively to avoid possible oversegmentation of characters. Finally, an algorithm based on a rating scheme is proposed to determine the cluster where the text pixels are classified. The proposed method is evaluated on the basis of recognition results instead of visual segmentation results. Comparative experimental results using a test set of 2684 characters are reported.

show abstract

Section: Resolution Enhancementmentioning

confidence: 97%

Section: Related Workmentioning

confidence: 99%

“…Before proceeding further, all feature components are scaled to the range [0, 1]. Additional details about the feature extraction process can be found in [9].…”

Section: Feature Extractionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Adaptive Fuzzy Text Segmentation in Images with Complex Backgrounds Using Color and Texture

Gllavata

Freisleben

2005

Lecture Notes in Computer Science

Self Cite

View full text Add to dashboard Cite

show abstract

“…This led to a plethora of methods that are surveyed by Jung et al (2004). The work conducted in the authors' workgroup includes proposals for text detection (Gllavata/Ewerth/Freisleben, 2004a) and text segmentation (Gllavata/Ewerth/Stefi/Freisleben, 2004;Gllavata/Freisleben, 2005) as well as a method for tracking moving text across several video frames (Gllavata/Ewerth/Freisleben, 2004b). The proposed text segmentation method was able to boost the word (character) recognition rate from 62% to 79% (76% to 91%) on a set of test images (Gllavata/Freisleben, 2005).…”

Section: Detection and Recognition Of Superimposed Textmentioning

confidence: 99%

Videana: A Software Toolkit for Scientific Film Studies

Ewerth¹,

Mühling²,

Stadelmann³

et al. 2009

Digital Tools in Media Studies

Self Cite

View full text Add to dashboard Cite

Within the research project "Methods and Tools for Computer-Assisted Media Analysis" funded by Deutsche Forschungsgemeinschaft, we have developed the software toolkit Videana to relieve media scholars from the time-consuming task of annotating videos and films manually. In this paper, we present the automatic analysis tools and the graphical user interface (GUI) of Videana. The following automatic video content analysis approaches are part of Videana: shot boundary detection, camera motion estimation, detection and recognition of superimposed text, detection and recognition of faces in a video, and audio segmentation. The GUI of Videana allows the user to subsequently correct erroneous detection results and to insert user-defined comments or keywords at the shot level. Furthermore, several research applications of Videana are discussed. Finally, experimental results are presented for the content analysis approaches and compared to the quality of human annotations.

show abstract

Text Segmentation in Complex Background Based on Color and Scale Information of Character Strokes

Wang

Gao

2007

Advances in Multimedia Information Processing – PCM 2007

View full text Add to dashboard Cite

Unsupervised Text Segmentation Using Color and Wavelet Features

Cited by 19 publications

References 11 publications

Adaptive Fuzzy Text Segmentation in Images with Complex Backgrounds Using Color and Texture

Adaptive Fuzzy Text Segmentation in Images with Complex Backgrounds Using Color and Texture

Videana: A Software Toolkit for Scientific Film Studies

Text Segmentation in Complex Background Based on Color and Scale Information of Character Strokes

Contact Info

Product

Resources

About