Speedy character line detection algorithm using image block-based histogram analysis

Academic institutions such as universities and technical colleges usually employ paper-based examinations and reports to evaluate the academic performance of students. Consequently, teachers expend considerable time and energy in the marking of such paper-based examinations. We are developing an automatic paper marking system geared towards reducing this paper-marking burden on teachers. To execute paper marking, handwritten character lines are extracted from examination papers, and then characters on those lines are recognized. In this paper, we primarily discuss how the character line is extracted from handwritten examination papers without ruled lines. The extraction of character lines from non-ruled papers is difficult because of the writing characteristic of students. Further, extraction accuracy is an important factor in character recognition performance. Conventional character line extraction algorithms for printed documents perform poorly on this problem. Furthermore, most proposed methods conduct tests using document images that include only character lines. In this paper we develop a less time-consuming algorithm for this task.