Optimizing graphics geometry using similarity-based clustering

摘要

A technique is described for clustering graphics objects in a graphic artwork based on affine variance. Geometry data associated with an input set of graphics objects can be processed to identify pairs of graphics objects that are affine variants of each other. The identified affine variants can then be clustered into one or more affine groups. Similar graphics objects in a given affine group can then be represented using a base geometry and a corresponding transformation matrix. Accordingly, the overall geometric complexity of the graphic artwork can be simplified thereby providing various benefits such as reducing file size and improving rendering performance.