Automatic extraction of node information for chemical molecular structure images

A chemical molecular image represents the connection relations among atoms, which uses nodes to stand for atoms of compound structure and uses edges to express connection bonds. The extraction of node information is focused on in the paper, and the main works include: (1) dividing nodes into two kinds in terms of node contents: intersection and endpoint; (2) identifying intersections through computing the deviation of pixels, which are lager deviation of others; and (3) extracting atomic contents of nodes after identifying nodes and merging them. Through implementing experiments for BMP images of chemical molecular structure, it shows that the accuracy of extracting atomic information of nodes by the method proposed is very high. The work lays foundations for automatic extraction information for chemical molecular structure images, and for chemical molecular structure images-based retrieval.