Dissertation > Excellent graduate degree dissertation topics show

Research on Video Text Extraction and the Application in Virtual Karaoke

Author: WangZuo
Tutor: ChenLinQiang
School: Hangzhou University of Electronic Science and Technology
Course: Computer Software and Theory
Keywords: Video Retrieval Text detection Video text positioning Text Segmentation Character recognition Background modeling
CLC: TP391.41
Type: Master's thesis
Year: 2011
Downloads: 33
Quote: 0
Read: Download Dissertation

Abstract


The video text contains rich semantic information, text extraction based on video content analysis, retrieval and other research fields play an important role. If the video text detection, segmentation, identified automatically understand the high-level semantic video retrieval is very valuable. Different from the ordinary scanning documents OCR recognition can be used directly to extract video text, there is a great difference in the video text attributes such as size, shape, color, and text in most cases also in complex background, these text extraction process hindered. How to take advantage of the characteristics of the video text, based on existing research to better text extraction, has become the focus of this study. This paper presents a comprehensive utilization of text color may exist several colors subtitles in the video, the edge of the geometric characteristics of the method. First detecting continuous multi-frame of the text position of the gradient method, and then the position is mapped to the original video frame, thus obtaining a large number of accurate and reliable text color information, based on these color information using GMM (Gaussian Mixture Model) color modeling, model in the subsequent frame extract text color layer, and update the model. According to the video with the redundancy of time, \The method of comprehensive utilization of the various features of the text, the large gap between the background and text color in the video can effectively extract the text. Still text in the video on the temporal and spatial redundancy characteristics, this paper presents a detection - method of tracking. First video frame edge detection to obtain the text area, and then to the edge of the text area bitmap matching feature tracking text, The refined test results quickly and efficiently locate text object. Text tracking also avoided for each frame segmentation, to identify and reduce the amount of computation. In the segmentation stage, lower the resolution of the video text, first with a multi-frame fusion method to enhance the text area and then further interpolation to enlarge the text. Design a virtual karaoke OK, the Kara OK video text extraction and person detection. In the Kara OK video text positioning, the use of a combination of wavelet transform and morphological method. Harr wavelet decomposition of the video frames after the opening, the closing operation decomposition subgraph, positioned in the maximum extent up in addition to the the oblique high frequency band of the background noise is mapped to the original video subtitle area. The method is not sensitive to the color, better positioning of the text of the karaoke OK. On character segmentation, selection of a single Gaussian background modeling background subtraction. Finally the extracted text and segmented characters with any selected scene for image fusion, the edge of the characters do fuzzy processing in order to achieve a better visual effect. The proposed two methods are chosen are several different types of video performance test, the experimental results show that these methods have a high detection performance, better able to extract the text in the different types of video.

Related Dissertations

  1. Moving target trajectory analysis based Intelligent Traffic Monitoring System,TP277
  2. Design and implementation of intelligent transportation system based on license plate recognition technology,TP391.41
  3. The Study of Moving Object Detection and Tracking Algorithm Based on Image Information,TP391.41
  4. Research and Realization of License Plate Character Recognition Algorithm Based on SVM,TP391.41
  5. Number of feature-based recognition of the banknote,TP391.41
  6. Image search on the web sensitive information in real time early warning system application,TP391.41
  7. Research on People Counting in Real-time Video Surveillance,TP391.41
  8. Research of Handwritten Chinese Character Recognition,TP391.43
  9. Video Retrieval for H.264 Compressed Video,TP391.41
  10. ICR - based banking supervision system design,F832.2
  11. Printed mathematical expressions Location Technology,TP391.41
  12. Based on MPEG-7 video database storage and retrieval technology,TP391.41
  13. Image Processing Based on Pattern Recognition and Its Application to License Plate Recognition,TP391.41
  14. The Research and Design for License Plate Recognition System Based on DSP,TP391.41
  15. Realize of Moving Object Detection Algorithm Designed Based on DM642,TP391.41
  16. Research on Apple Lesion Retrieval Based on Video Shots,TP391.41
  17. Instrument dial numeric characters automatic identification technology research,TP391.41
  18. DSP-based dynamic multi- digital instrument Character Recognition System,TP391.41
  19. Research on Content-based Video Retrieval Keytechnologies,TP391.41
  20. Digital image processing in the vehicle license plate recognition research,TP391.41
  21. Based on sparse representation of off-line handwritten Chinese character recognition,TP391.1

CLC: > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device
© 2012 www.DissertationTopic.Net  Mobile