Detecting formula based on Stroke Width Transform for online Chinese examination question retrieval

More and more students prefer to submit the questions captured by cell phone to the image recognition and retrieval system to find the answers when they are self-taught. In this paper, we propose a new method to detect formulas in text lines captured by cell phone in natural conditions for online Chinese examination question retrieval. First, we use the stroke width transform algorithm to get the initial text lines. Second, we take an improved measure to merge these initially scattered text lines into principal text lines, which can be used for the following operation. Next, characters in the text lines are segmented by projection method and marked as Chinese character or non-Chinese-character automatically by some features. If the neighboring characters match the limit conditions, they will be detected as a formula and be extracted. Since some formulas are higher than the localized text lines which they belong to, an extension operation will be done to the extracted formula in the vertical direction to get the complete formula. We tested our method on some data sets, and the result shows that our method is satisfactory for various subjects.