Detecting and Localizing Visual Inconsistencies in Web Applications

Failures in the presentation layer of a web application can negatively impact its usability and end users' perception of the application's quality. The problem of verifying the consistency of a web application's user interface across its different pages is one of the many challenges that software development teams face in testing the presentation layer. In this paper we propose a novel automated approach to detect and localize visual inconsistencies in web applications. To detect visual inconsistencies, our approach uses computer vision techniques to compare a test web page with its reference. Then to localize, our approach analyzes the structure and style of the underlying HTML elements to find the faulty elements responsible for the observed inconsistencies.