DOCUMENT SKEW DETECTION — A NOVEL APPROACH
In this paper, we propose a novel skew estimation technique for binary document images based on Boundary Growing Method (BGM), thinning and moments. BGM helps in extracting the text line blocks from the document. Thinning1 is performed to fit the best line for extracted text line blocks. Further, skew is computed for thinned line using second order moments. Several experiments have been conducted on various types of documents such as documents containing south Indian languages, English documents, journals, text with picture, noisy images, and document with different fonts and resolutions, to reveal the robustness of the proposed method. Based on the experimental results we have realized that the proposed method outperforms existing methods both in terms of mean and standard deviation.