Improve The Character Detection System Based On Feature Extraction Algorithm

The character recognition is the major important part in the area of document analysis. Character Recognition could be evaluated on printed text and handwritten text. Printed texture could be from a good quality image. In this research work, we implemented in the OCR approach to improve the recognition of character with Classification approach. We work on filtration techniques to improve the pixel quality of the punjabi character images. In bilateral filter works binary pixels could be close to single another, i.e occupies nearby locations they could be same to another that is having close values possibly in a perceptually meaningful manner. An inverse filter is used to recompense for the effect of unwanted structure filtering of signals. The quality of the character is improving and increase the accuracy rate due to presence of characters in the image. To enhance the accuracy with the help of a classification approach i.e FFNN. In this approach work in two phases i.e Training State and Testing State. In training State to evaluate the performance based on trained features in the punjabi character image. We decided the epoch value is 100 means, train the punjabi characters in this given reputations. After that train the features, then simulation model for analysis the features in testing stage. From the existing research, it is clear that detect the punjabi characters are accuracy better than text detection systems in terms of security, accuracy, performance and image efficiency and quality. Existing research proposed optical character recognition systems that are detecting the punjabi character, but doesn’t improve the accuracy parameter. Tool used for simulation is MATLAB. Keywords— Optical Character Recognition (OCR), Classification, Digital Image process, Feed Forward Neural Network(FFNN),Support Vector Machine(SVM).