Table of Contents
Partial Shape Similarity
0. Motivation
1. Edge Detection
2. Edge Grouping
3. Extraction of Contour Parts
4. Recognizing Shapes by Parts
5. Results
Structural Shape Similarity
0. Introduction
1. Shape Segmentation
2. Skeleton Computation
3. Skeleton Pruning
4. Recognizing Shapes by Structures
Contour Grouping Based on Local Symetry
0. Introduction
1. From Edges To Contours
2. Center points
3. Reference Shape Model
4. Particle Filter
5.1. Experimental Result 1
5.2. Experimental Result 2
5.3. Experimental Result 3
5.4. Experimental Result 4
5.5. Experimental Result 5
5.6. Experimental Result 6
5.7. Experimental Result 7
5.8. Experimental Result 8
6. Conclusion
7. Contour Grouping Video
Edge Grouping Extraction of Contour Parts Extraction of Contour Parts

Extraction of Contour Parts



An elephant contour undergoing DCE
This animation shows an elephant contour, both before and after DCE; the common points between the original contour and the evolved contour; and some of the separate contour parts demarcated by the DCE points.

Click the image for the full size animation.
Discrete Curve Evolution

Discrete Curve Evolution (DCE) was an algorithm developed by Drs. Longin Jan Latecki and Rolf Lakaemper for simplifying contour curves.The power of DCE is that it preserves the outline of a shape by preserving the most visually critical points.

[Read More On DCE]
[Convexivity Rule...]
[Shape similarity measure...]

The next step in building a useful representation is to use Discrete Curve Evolution (DCE) to find points of visual significance. DCE is built upon the idea that all digital curves are actually polygons with possibly large numbers of vertices. Through DCE, the set of only visually important vertices is extracted from the set of all points on the contour.

This set of important points is used to decompose the original contour into useful parts for comparison. Each DCE point can serve as an end point for a visually significant contour part. Because DCE decomposes similar shapes in a similar manner, even through noise and distortion, the ends of partial contours extracted will roughly correlate to contour ends of similar shapes. We choose the longest contour part of the query image as the query contour.