Reducing Redundant Checking for Visual Regression Testing

The regression testing method called Visual Regression Testing checks whether application screens display correctly without any presentation failures. This method uses image comparison tools that detect differences in screen elements, e.g. the disappearance of or change in the position of a button,...

Full description

Saved in:
Bibliographic Details
Main Authors: Adachi, Yu, Tanno, Haruto, Yoshimura, Yu
Format: Conference Proceeding
Language:English
Subjects:
Computer vision
group the same differences
Navigation
reduce redundant checking
Software
Software engineering
Technological innovation
Testing
visual regression testing
Visualization
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The regression testing method called Visual Regression Testing checks whether application screens display correctly without any presentation failures. This method uses image comparison tools that detect differences in screen elements, e.g. the disappearance of or change in the position of a button, by comparing images of the correct screen and the target screen by using computer vision techniques. It can detect differences in all the screens of an application automatically, but the tester needs to confirm all the differences detected in the screens. On the other hand, the application screen is often composed of areas such as the header, footer, navigation and content areas. Since the header, footer and navigation are displayed at fixed positions on a screen and commonly exist on many screens, if a difference is detected in them, the same difference will be detected in many other screens. As a result, we need labor to repeat the same confirmation for screens containing the same differences. Here, we propose a method to reduce the labor in making such confirmations by analyzing all the screen comparison results to identify the same differences so that the tester only has to confirm one screen comparison result as a representative. We confirmed the effectiveness of our method on real application screens and confirmed that it reduces the number of differences that the tester needs to confirm by about 60%.
ISSN:2640-0715
DOI:10.1109/APSEC.2018.00106