Detecting Characters From Digital Image With Chinese Text Based on Grid Lines
Anshuang Sun, Xiwen Zhang · 2019
For neatness, many Chinese and international students write on paper with grid lines in horizontal and vertical directions. The lines have a few styles, such as composition format, post format and so on. The approaches of Chinese text detection in natural scenes focus on complex and changeable objects, paying less attention to grid lines. Therefore, the results of their application are mostly unsatisfactory. In Chinese text image with grid lines, characters are surrounded by similar rectangular boundary lines. This paper proposes an approach based on grid lines to detect characters. The approach can effectively avoid the omissions and overlaps that easily occurs in other approaches. The image is firstly converted into a binary edge one using adaptive Canny edge detection. Horizontal and vertical lines are extracted from the binary edge image using Hough transform. Rectangles bounding characters are identified from the transformed lines. The experimental results show that this presented approach can accurately detect characters from different formats of Chinese text images with the grid lines.