The practical Application of Knowledge Discovery to Image Data: A Practitioners View in The Context of Image Analysis
 |
Prof. Frans Coenen |
Department of Computer Science The University of Liverpool Liverpool, UK |
|
|
Abstract: Knowledge Discovery in Images (KDI) is concerned with the discovery of hidden information in image data of all kinds. The actual image mining element is well understood. The challenge is the end-to-end process of knowledge discovery in images from translating the input data into a form whereby it can be mined to understanding the result. A review is presented of image mining in the wider context, both 2D and 3D, in terms of the mechanisms where by image data can be translated into an eventual feature vector format to which established data mining techniques, of many kinds, can be applied. A number of alternative representations are considered: graph based, point series based, histogram based and others. The ideas presented are illustrated using a number of applications including Magnetic Resonance Image analysis (2D and 3D), Google Earth satellite image interpretation for population estimation, retina image diagnosis (2D and 3D) and 3D surface analysis in the sheet metal forming industry. |
|