Table frame line detection in low quality document images based on Hough transform

Table detection is of importance in the field of document images analysis and processing, especially table frame line detection. Although a great success has been achieved for high quality images during the past decade, table detection in low quality images still remains a challenge. To address this problem, we proposed a neoteric method to detect table frame line automatically in low quality document images. Firstly, Radon transform is adopted to detect skew of document images and then correct it. Secondly, run length smoothing algorithm (RLSA) is used to extract the lines longer than a predefined threshold. Thirdly, we locate table regions according to table features and detect frame lines of the detected tables using Hough transform method. The experimental results show that this method could obtain a better performance even in the low quality document images compared to the conventional method.