OCR From a Web Camera

10,000 3,000

Topic Description

 ALL listed project topics on our website are complete material from chapter 1-5 which are well supervised and approved by lecturers who are intellectual in their various fields of discipline, documented to assist you with complete, quality and well organized researched materials. which should be use as reference or Guild line...  See frequently asked questions and answeres

This project addresses the problems of performing Optical Character Recognition on images acquired
from web cameras. The overall aim of the project was to study the limitations of OCR engines, to analyse the complexities of images acquired from web cameras and to produce a program which improves OCR performance on such images. The aim can be refined into several objectives:
• Understand OCR limitations
• Understand the image complexities when acquisition is performed by a web camera
• Collect a dataset of images demonstrating image complexities
• Obtain an OCR package
• Evaluate the success of the OCR package on the dataset
• Define the scope of the program to be developed
• Research and select or design an algorithm to improve the suitability of images for OCR purposes
• Implement the algorithm
• Test the algorithm on a large dataset
• Refine the algorithm if necessary and retest
• Evaluate the results against an OCR package
The report evaluates an OCR package, ‘Page Cam’, and highlights the complexities of the problem; such as image skew, perspective distortions, image resolution, textual mediums, illumination variances and lens effects. The scope of the project is defined, as improving the suitability for OCR of skewed images. A literature review is performed, and a program is developed which attempts to deskew images and remove illumination effects. The program corrects images containing non-textual information, such as illustrations, skew in the range ± 90o and uniform or non-uniform light effects. The program is tested and several problems surface, related to the removal of light effects and the allowance of non-textual
information. Possible solutions are researched and then implemented, before testing resumes. Testing
revealed promising results, such that ‘Page Cam’ OCR accuracy improved by thirty-eight percent when
additionally using the developed program.
The project is evaluated and several program improvements are discussed, such as a rotation algorithm
which reduces aliasing effects, and an enhanced textual segmentation method. Future areas of work are
also reviewed, including the improvement of perspectively distorted images, the allowance of a range of
textual mediums and image resolution enhancement. The implemented program is compared with
previous attempts, before the project concludes by discussing the chosen development methodology and
reflecting on the project experience.