Technologies are generally described for a system to extract description of reference numerals in images and facilitate keyword-based search in images. In various examples, the system may include one or more databases, a computer readable memory, and one or more processors. The system may be configured to extract one or more reference numerals from an image, and identify and extract corresponding description of the one or more reference numerals from a description document corresponding to the image. The system may be further configured to extract text from the images, and store the images in a database with the extracted data, i.e., text, reference numerals, and corresponding descriptions. The system may be further configured to receive an input query intending to search images related to a search logic of the input query, search a database to identify an image including either of text or a reference numeral having a corresponding description that corresponds to the search logic of the input query, and render the identified image via a display device executing the output interface.