A general image skew detection approach with a bio-inspired mechanism

This paper proposes a simple but effective bio-inspired approach for skew detection of different kinds of images including nature images, documents, paintings etc., which requires no parameter tuning. Without relying on local visual cues, the proposed approach exploits the global statistics of an image. Orientation perceptron is designed in the inspiration of biological simple cells for local orientation detection. Statistics of their outputs are processed and then transferred by orientation transform into a curve whose maximum value is used to determine the skew angle. The effectiveness of the proposed approach is verified by experiments.