doi:10.1016/j.imavis.2006.06.017
Copyright © 2006 Elsevier B.V. All rights reserved.
A shape interpolation technique based on inclusion relationships and median sets
aArtificial Intelligence Laboratory, Facultad de Informática, Universidad Politécnica de Madrid, 28660 Boadilla del Monte (Madrid), Spain
bComputer Science Department, Universidad de Concepción, Chile
Available online 7 November 2006.
References and further reading may be available for this article. To view references and further reading you must
purchase this article.
Abstract
Some image processing and analysis applications require performing image interpolation. This paper focuses on interpolation approaches that treat the shapes and the structures of binary images. A summary of some interpolation methods is presented, and their behavior concerning inclusion relationships and homotopy issues is studied. An inclusion relationship property that considers these aspects is introduced in this work. Furthermore, a complete technique based on this property in a recursive manner is presented. The paper shows that such a property can improve shape interpolation results in a relatively easy manner, particularly those concerning inclusion relationships between shapes. Several experimental results are provided.
Keywords: Image processing; Interpolation; Shape interpolation; Mathematical morphology; Median set
Fig. 1. Sequence of interpolated slices (and their homotopy trees) obtained using interpolation based on Hausdorff distance with convex mask.
Fig. 2. Sequence of interpolated slices (and their homotopy trees) generated by median set-based interpolation method.
Fig. 3. Sequence of interpolated slices (and their homotopy trees) generated by the interpolation function method. (Note that there are two pores in image (d) and three grains in images (i) and (j).)
Fig. 4. Inclusion property illustrated.
Fig. 5. Inclusion property and homotopy issues.
Fig. 6. General interpolation algorithm. (Functions matching and displace are described later.)
Fig. 7. Separation of outer filled CCs.
Fig. 8. First matching criterion: (a) a slice with one CC A, indicating its MSP and its radius, λA; (b) a second slice with CC B; and (c) the proximity zone of A, δλA (A) (in gray).
Fig. 9. Matching algorithm.
Fig. 11. Alignment and interpolated set:(a) Input sets A and B and their respective MSP points; and (b) Aligning of both input sets and interpolated set.
Fig. 13. Extreme case in which inclusion is desired.
Fig. 14. Position of holes in interpolated shapes: (a) Shape 1; (b) Shape 2; (c) Aligned outer CCs and interpolated outer grain; (d) Position of holes; (e) Aligned holes; and (f) Interpolated result.
Fig. 15. Selecting new locations for inner structures.
Fig. 16. Displacement algorithm.
Fig. 17. Interpolated results for the extreme example of Fig. 13.
Fig. 18. Simple example: interpolation of a circle and a square. Slices 1 and 2 are images (a) and (i), respectively. Images (b) to (h) are interpolated results. Image (j) represents the homotopy tree for all slices.
Fig. 19. Interpolated results 2 (compare with Fig. 2). Slices 1 and 2 are images (a) and (g), respectively. Images (b) to (f) are interpolated results. Image (h) represents the homotopy tree for all slices.
Fig. 20. Interpolated result 3 (compare with Fig. 3). Slices 1 and 2 are images (a) and (g), respectively. Images (b) to (f) are interpolated results. Image (h) represents the homotopy tree for all slices.
Fig. 21. Interpolation results 4: interpolation of nested CCs. Slices 1 and 2 are images (a) and (g), respectively. Images (b) to (f) are interpolated results. Image (h) represents the homotopy tree for all slices.
Fig. 22. Interpolation results 5: interpolation of slices with different homotopy (different number of grains). Slices 1 and 2 are images (a) and (h), respectively. Images (b) to (g) are interpolated results.
Fig. 23. Interpolation results 6: complex case with different homotopy and nested CCs . Slices 1 and 2 are images (a) and (i), respectively. Images (b) to (h) are interpolated results.
Table 1.
Distance table between CCs of Fig. 10
