|
|
Bimonthly Since 1986 |
ISSN 1004-9037
|
|
|
|
|
Publication Details |
Edited by: Editorial Board of Journal of Data Acquisition and Processing
P.O. Box 2704, Beijing 100190, P.R. China
Sponsored by: Institute of Computing Technology, CAS & China Computer Federation
Undertaken by: Institute of Computing Technology, CAS
Published by: SCIENCE PRESS, BEIJING, CHINA
Distributed by:
China: All Local Post Offices
|
|
|
|
|
|
|
|
|
|
Abstract
The identification and recognition of scene-text remains an open and challenging subject in the field of computer vision. This is due to the fact that text often covers just a tiny portion of an image and has a backdrop that is not uniform. The scene-text detection and identification problem are made more difficult by factors such as noise, blur, occlusions, and reflections. As a result, insufficient accuracy and excessive time complexity are produced as a result of this challenge. In order to locate the accentuated edges in the scene-text image, the Sobel and Gaussian Edge Detectors are converted into Energetic Edge Detectors. In order to accurately localise the scene-text, the updated Local Directional Number descriptor on the Energetic Edge Information drives a course of action forward. The innovative algorithm that identifies the text characters of localised scene-text is able to account for these phenomena. The recognised output is supplied in the form of characters from the English language. The Energy enriched SOM that is empowered by the Histogram of Oriented Gradients and the Symlet Transform is the primary factor responsible for the success of the procedure. The output of this highly accurate scene-text recognition system may be coupled to simple activities that include image processing.
Keyword
scene-text recognition, Neural Networks, self-organising map, Hog Descriptor
PDF Download (click here)
|
|
|
|
|