A Novel Method for Efficient Text Extraction from Real Time Images with Diversified Background using Haar Discrete Wavelet Transform and K-Means Clustering

The proposed system highlights a novel approach of extracting a text from image using two dimensional Haar Discrete Wavelet Transformation and K-Means Clustering. As the commercial usage of digital contents are on rise, the requirement of an efficient and error free indexing text along with text localization and extraction is of high importance. Majority of the previous research work on text extraction has focused on scene text, uniform background, and extensive use of wavelet domain and frequent usage of only grey-scale image as input. The extensive in-depth testing of such approach will lead no not-so-satisfactory results if the image type, non-uniform background, different text orientation, different languages are introduced. The proposed system has broader scale of consideration of input image with much complicated backgrounds along with consideration of sliding windows. For much accuracy, morphological operation is included to accurately distinguish the text and non-text area for better text localization and extraction. The experimental result was compared with all the prior significant work in text extraction where the results show a much robust, efficient, and much accurate text extraction technique. Keyword: Text Extraction, Haar, Discrete Wavelet Transform, K-Means Clustering, Morphological Operations