Extraction of Text Regions in Natural Images through Boosting

In this paper, we first transform all the images into a gray level in the pre-processing step, and using the Wiener method on the images in the next step, we denoise all the images. In the next step, which is feature extraction of a binary texture pattern, we apply the denoised images to the Gist-type texture algorithm, and prepare the features to classify the images into text and non-text categories, and apply this set of features to the AdaBoost classification. We use the AdaBoost method as the best and most effective method for boosting in this type of implementation with a 90% accuracy percentage. Finally, to validate the answer and eliminate the stochastic conditions in the training and experimental phases, we use the conventional 10-fold cross-validation method

Extraction of Text Regions in Natural Images through Boosting Keywords:

Wiener , feature extraction , Gist , classification , AdaBoost