A Parsing Scheme of Mind-Map Images

Bo Wang, Ju Zhou, Bailing Zhang

Abstract: At present, there are a large number of mind-map images generated by different software on the Internet. These images contain rich knowledge that has been summarized by people for learning or teaching. In order to exploit knowledge and reconstruct the mind-map, we propose a scheme to parse the image of mind-maps. Firstly, we apply text detection and deep learning-based OCR models to extract the text content and the corresponding position information, and then use the line segment detection model to detect the line segments which represent the structural relationships between text blocks. The detected line segments will be matched with the text blocks with the image features of the mind map. The results are utilized to analyze the parent-child relationship between the text blocks. Experiments have been conducted to illustrate the reconstruction of mind-map with the parsing of the corresponding images.