Skip to Main Content
 

Global Search Box

 
 
 
 

Files

File List


Thesis.pdf (1.48 MB)

ETD Abstract Container

Abstract Header

An Investigative and Goal driven Workbench for Text Extraction and Image Processing

Tumu, Sudheer
http://rave.ohiolink.edu/etdc/view?acc_num=osu1376930066

Abstract Details

2013, Master of Science, Ohio State University, Computer Science and Engineering.
Text data present in images and video provide useful information for indexing, annotation and structuring of images [1]. However, automatic extraction of text is extremely challenging because of variations in the source images such as contrast, complexity of background and as well as variations in the text to be extracted in style, size and orientation. This requires systematic experimentation where experiments are recorded, results are saved etc. Hence an “experimental workbench” that consists of various basic image processing and data analysis tools is needed to conduct an experiment or a series of experiments to achieve goals such as text extraction, basic image processing and to save intermediate/final results. This document presents the design and implementation of an experimental workbench that provides a collection of basic image processing and text extraction tools that an individual or an organization can use to perform various tasks such as extracting text from an image or a video. The transformations provided in the workbench are image to image transformations such as smoothing, dilation and erosion; image to text transformation such as optical character recognition (OCR); and text to text transformation such as fuzzy matching the extracted text using OCR, with an existing knowledge database to improve accuracy of extracted text. In addition to that, the workbench also provides support for automation and orchestration of existing tools. Users can create custom tools/transformations by combining existing tools and save intermediate results as checkpoints that can be used to roll back if necessary. The workbench was used to build an online library catalog by extracting book titles from a video stream of book spines. A custom transformation was created to perform this task, which is named as `hill-climbing’ that automates a series of basic image processing and text extraction tools. The video stream was recorded by holding the camera facing book spines and walking across the book shelf. The main contributions of our work are thus: providing an integrated collection of image processing operations, designing and developing a workbench of a basic image processing and text extraction tools from image or video, and using the proposed workbench to build an online library catalog from a video stream of book spines.
Rajiv Ramnath (Advisor)
Jay Ramanathan (Committee Member)
66 p.
Computer Science

Recommended Citations

Citations

  • Tumu, S. (2013). An Investigative and Goal driven Workbench for Text Extraction and Image Processing [Master's thesis, Ohio State University]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=osu1376930066

    APA Style (7th edition)

  • Tumu, Sudheer. An Investigative and Goal driven Workbench for Text Extraction and Image Processing . 2013. Ohio State University, Master's thesis. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=osu1376930066.

    MLA Style (8th edition)

  • Tumu, Sudheer. "An Investigative and Goal driven Workbench for Text Extraction and Image Processing ." Master's thesis, Ohio State University, 2013. http://rave.ohiolink.edu/etdc/view?acc_num=osu1376930066

    Chicago Manual of Style (17th edition)

osu1376930066
434
© 2013, all rights reserved.
This open access ETD is published by The Ohio State University and OhioLINK.