Text Line Detection Based on Mutation Analysis

A text line detection method based on wavelet transformation and mutation analysis is proposed in this paper. First the character density image is acquired from the input document image by the project function with strips, and then the density image is analyzed by wavelet transformation where the Gaussian or its derivative function is used as the wavelet basis. The local peak positions indicate the line texture which will be continuous and single pixel. Because of the statistical properties of project function and wavelet transformation of mutations signal detection, this algorithm is not sensitive to noise and warped distortions. Experiments show that this algorithm is accurate and robust.