Reading 15 : IStraw

Citation:

Xiong, Yiyan, and Joseph J. LaViola Jr. "Revisiting ShortStraw: improving corner finding in sketch-based interfaces." Proceedings of the 6th Eurographics Symposium on Sketch-Based Interfaces and Modeling. ACM, 2009.

Publication Link 

 

Summary:

This paper first reviews the Shortstraw algorithm, and then identifies its limitations and suggestions improvements. It also introduces curve detection, and provides experimental proof for the massive improvement obtained.

 

Discussion:

We mainly need to discuss the 6 limitations that occur in ShortStraw, and how iStraw overcomes them.

 

1. Initial Straws

Problem : Straw fails to compute the straw for the first 3 and last 3 points, which can turn out to miss corners in some cases. 

Solution : iStraw introduces formalae to compute the straws for these points too.

2. Timing information

Problem  : Shortstraw does not make any use of the stroke speed, which can actually be helpful in finding corners 

Solution. Use speed maxima to identify corners.

3. Consecutive False Corners Avoidance

Problem: Shortstraw tends to delete some true corners, after which the colinearity test fails drastically.

Solution: Use a two-pass co-linearity test, with second Dpass having more relaxed thresholds.

4. Dynamic Threshold

Problem : Shortstraw does not take into account the length of line segments, while determining corners. This is an important distinguishing factor is some cases.

Solution : In the second pass of colinearity test, a dynamic threshold based on length of the segment, is introduced.

5. Impact of Sharp noise

6. iStraw introduces curve detection by treating every prospective corner as a point where two rays coverage, both originating from a fixed shift distance from the point in question. For a line, the angle formed a this point remains fairly constant, while it changes drastically for a curve.