Detection of the global text skew with initial skew rate

Old documents made on a typewriter represent significant part of our technical and cultural inheritance. Their digitalization plays an important role. The paper proposes the algorithm for identification of the global text skew with the help of initial text skew. It is based on connected components. These elements are established by joining convex hulls, which encircled each text element. Furthermore, initial text skew is identified. Then, connected components are enlarged with directed morphological erosion which is determined by the initial text skew. At the end, the biggest connected component is extracted. Its orientation determines the global orientation of text document. Efficiency and correctness of the algorithm are verified by testing on custom dataset.