Algorithmic Optimization of Scalable Vector Graphics Through Structural Simplification and Transform Consolidation

Abstract

Scalable Vector Graphics (SVG) files are widely used on the web because of their smaller size compared to their raster counterparts. However, SVG files that are authored from vector suites might not have the most optimized representation. They contain significant data overhead and redundancies mostly arising from verbose metadata and non-optimized path definitions. To address these issues this work introduces a series of algorithms that target key areas that could yield size improvements. The algorithms work to decrease complex path definitions to much simpler representable SVG commands. To address attribute verbosity, the paper proposes a method that takes in sequence of transformation commands and consolidate them into one single transformation matrix. Additionally, it also tackles embedded assets which have applied filters by pre-multiplying them ahead of time, allowing for the total removal of filter tags. After applying these methods, the results have shown noticeable improvements of around 8.41"\%"  on some average. Although, the potential for compression and optimization depends highly on those files that are applicable to these transformations. This work has combined methods from areas such as computational geometry and rendering into SVG and has demonstrated that there is room for improvement for more efficient representations within SVG without introducing a new vector image standard.