Efficient and robust image descriptor for GUI object classification

Graphical User Interface (GUI) object classification is essential for image-based software automation tools. The challenges posed by GUI object classification are significantly different from those in natural image classification. In this paper we present a novel image descriptor developed specifically for GUI objects; it is robust to various changes in the appearance of GUI objects, such as various screen resolution, 'skin', as well as various operating system related. We use this image descriptor with Support Vector Machine classifier, and experimentally show the descriptor robustness to the above transformations, and its superior performance compared to existing image descriptors. © 2012 ICPR Org Committee.