Automated Visual Perception-Based Web Browser Rendering Results Comparison with Multi-part Fragment Image Matching

摘要

A difference measurement method for similar web pages is proposed. Displaying same html documents on a different browsers or different browser versions may produce different results, when compared as images. They may be visually indistinguishable or may visually differ. There are many factors causing that differences. There are ones as advertisements introduced or changed randomly by some sites, which should be handled accordingly. The proposed method automatically register different part of compared images and employ VDP method to produce visual difference map and compute difference coefficients. Obtained results shows, that it is possible properly compare web pages even if part of them change sizes causing translation of the resulting content. Proposed system was realised as a part of the automated testing environment suited to perform regression tests for browser engines.