A System for Recognizing and Beautifying Low-level Sketch Shapes Using NDDE and DCR

Sketching has been identified as a natural means for human interaction and thus has become commonly incorporated into various user interfaces. Current low-level sketch recognizers have produced good accuracy but recognize only a small set of basic shapes. We propose a low-level sketch recognition and beautification system that uses a hierarchical approach that is capable of recognizing eight primitive shapes, along with complex fits, with preliminary recognition rates around 98.8%. These accuracy rates are comparable to current stateof-the-art recognition systems which recognize a lesser number of primitives. Furthermore, we introduce two new metrics, normalized distance between direction extremes (NDDE) and direction change ratio (DCR), which help aid in distinguishing between polylines and other low-level primitives.