Knowledge extraction from images

• Main Author: Wang, Yuqi
• Other Authors: Mao Kezhi
• Format: Final Year Project
• Language: English
• Published: 2016
• Subjects: DRNTU::Engineering
• Online Access: http://hdl.handle.net/10356/67918
• Institution: Nanyang Technological University

Knowledge extraction is a process of knowledge creation from different types which are structured and unstructured of sources. Images belong to the group of unstructured sources. The extracted knowledge must be solid data that is in a readable format and interpretable format by machine or a given program used by end users achieving their purpose. Knowledge extraction from images aims to gain valuable information satisfying users’ requirements. With the development of technology, especially for Internet Industries, online image searching engines has arisen and improved for several years and led online searching technology to a new level. However, the precision and efficiency will be the most important parts for online image searching and needed to find ways to innovate and develop constantly. According to some existing machine learning theories and image processing principles, a refined search will be shown in this report for enhancing image search engines to think more like a human. The basic concept of such a thinking way is actually to combine two sub-steps of the process for extraction, one is extracting information and analysis from the image itself, the other one is extracting from text part tagged to the image, such as short descriptions or titles. The verification results demonstrate that the refined system will have an ideal and effective outcome and can be applied to real cases.