Edge detection and ridge detection with automatic scale selection
read more
Citations
Contour Detection and Hierarchical Image Segmentation
Computer Vision: Algorithms and Applications
Muliscale Vessel Enhancement Filtering
Feature Detection with Automatic Scale Selection
An unbiased detector of curvilinear structures
References
A Computational Approach to Edge Detection
Marching cubes: A high resolution 3D surface construction algorithm
Scale-space and edge detection using anisotropic diffusion
Theory of Edge Detection
Characterization of Signals From Multiscale Edges
Frequently Asked Questions (9)
Q2. What are the future works in "Edge detection and ridge detection with automatic scale selection" ?
A more extensive description can be found in ( Lindeberg 1996 ), including further theory, motivations, algorithmic descriptions, as well as more detailed discussions about the relations to previous work. These suggested principles may be far more general than the actual implementations presented so far.
Q3. What scales are used to compute the di erential descriptors in the edge?
The di erential descriptors in the edge de nition (4) (rewritten in terms of partial derivatives in Cartesian coordinates) are computed at a number of scales in scale-space.
Q4. What is the scale-space edge detection scheme?
The scale-space edges have been drawn as three-dimensional curves in scale-space, overlayed on a low-contrast copy of the original grey-level image in such a way that the height over the image plane represents the selected scale.
Q5. How is the scale-space representation of a given signal obtained?
The scale-space representation L of a given signal f is obtained by convolving f by Gaussian kernels g ofvarious widths t = 2 (Witkin 1983, Koenderink 1984, Florack et al. 1992, Lindeberg 1994).
Q6. What can be used to express a large number of visual operations?
The output from these operators can be used as a basis for expressing a large number of visual operations, such as feature detection, matching, and computation of shape cues.
Q7. What is the simplest measure of edge strength?
Based on the -parameterized normalized derivative concept in (2), the authors shall here consider the following two di erential expressions:G normL = t (L2x + L2y); T normL = t3 (L3x Lx3 + 3L2x Ly Lx2y+ 3Lx L 2 y Lxy2 + L 3 y Ly3):The rst entity, the square gradient magnitude, is the presumably simplest measure of edge strength to think of.
Q8. What is the result of applying the integrated edge detection scheme to real-world images?
a polygon approximation is constructed of the in-tersections of the two zero-crossing surfaces of Lvv and @t(E norm) that satisfy the sign conditions Lvvv < 0 and @t(E norm) < 0. Finally, a signi cance measure is computed for each edge by integrating the normalized edge strength measure along the curveG( ) = Z (x; t)2 q (G normL)(x; t) ds;T ( ) = Z (x; t)2 4 q (T normL)(x; t) ds:Fig. 2 shows the result of applying this scheme to two real-world images.
Q9. What is the way to express these?
A particularly convenient framework for expressing these is in terms of multi-scale di erential invariants or singularities of these.