Rozpoznávání textu a technologie OCR
Loading...
Downloads
3
Date issued
Authors
Chaloupka, Milan
Journal Title
Journal ISSN
Volume Title
Publisher
Vysoká škola báňská - Technická univerzita Ostrava
Location
Signature
Abstract
This thesis solves the problem of defects and imperfections in the input materials for the OCR and describes possible solutions for issues of recognition of printed text. The thesis is a description of the basic algorithms for the adjustment of raw materials, as algorithms for edge detection in image, color segmentation in image and smoothing produced poor image. The thesis is also an assessment of the algorithms and compare the results obtained. Finally, the issue is explained recognition of curved text and text in circle with the possibility of solving this problem.
Description
Import 29/09/2010
Subject(s)
OCR, RGB, grayscale image, thresholding, global thresholding, Otsu thresholding, p-percentage thresholding, edge detection, Prewitt operator, Sobel operator, Laplacian of Gaussian, invariants, Fourier descriptors, moments, normal moments, central moments, Hu´s seven moment invariants, Zernike moments, Pseudo-Zernike moments, text in circle, curved text, JPG, PDF, C#