Text extraction using edge detection and morphological dilation
Shutao Li, James Tin-Yau Kwok · 2005
In this paper, a method used to extract text regions from document images is proposed. It first finds edges by computing the gradient of each location. Then long edges, such as lines and isolated edges which are often due to noise, are removed. Finally, this edge map is morphologically dilated. Experimentally, the proposed method is robust to noise, skew and text orientation.