Difference between revisions of "Image processing"

From Helpful
Jump to: navigation, search
m (Created page with "==Feature detection== Where features tend to include: * Points - * Blobs - smooth areas that won't (necessarily) be detected by point detection. Their approximate centers may...")
 
m
Line 1: Line 1:
 +
{{stub}}
 +
 
==Feature detection==
 
==Feature detection==
  
Line 21: Line 23:
 
** has a well-defined position
 
** has a well-defined position
 
** preferably quite reproducible, that is, stable under relatively minor image alterations such as scale, rotation, translation, brightness.
 
** preferably quite reproducible, that is, stable under relatively minor image alterations such as scale, rotation, translation, brightness.
** useful in their direct image context - endpoints, corners, intersections
+
** useful in their direct image context - corners, endpoints, intersections
  
 
* Region of interest
 
* Region of interest
Line 39: Line 41:
 
-->
 
-->
  
<!--
 
 
===Edge detection===
 
===Edge detection===
 +
<!--
  
 
In theory:
 
In theory:
Line 51: Line 53:
 
* edge orientation
 
* edge orientation
 
* gives edge magnitude (lets us filter the most significant)
 
* gives edge magnitude (lets us filter the most significant)
 
  
  
 
In sampled raster images, there is always noise
 
In sampled raster images, there is always noise
 +
-->
 +
* Canny [http://en.wikipedia.org/wiki/Canny_edge_detector]
 +
* Canny-Deriche [http://en.wikipedia.org/wiki/Canny_edge_detector#Conclusion]
 +
* Differential [http://en.wikipedia.org/wiki/Edge_detection#Differential_edge_detection]
 +
* Sobel [http://en.wikipedia.org/wiki/Sobel_operator]
 +
* Prewitt [http://en.wikipedia.org/wiki/Prewitt]
 +
* Roberts Cross [http://en.wikipedia.org/wiki/Roberts_Cross]
  
  
-->
+
===Interest point / corner detection===
 +
* Harris operator [http://en.wikipedia.org/wiki/Corner_detection#The_Harris_.26_Stephens_.2F_Plessey_corner_detection_algorithm]
 +
* Shi and Tomasi [http://en.wikipedia.org/wiki/Corner_detection#The_Shi_and_Tomasi_corner_detection_algorithm]
 +
* Level curve curvature [http://en.wikipedia.org/wiki/Corner_detection#The_level_curve_curvature_approach]
 +
* SUSAN [http://en.wikipedia.org/wiki/Corner_detection#The_SUSAN_corner_detector]
 +
* FAST [http://en.wikipedia.org/wiki/Corner_detection#The_FAST_feature_detector]
  
  
==Object detection==
+
===Blob detection ===
 +
* Laplacian of Gaussian (LoG)
 +
* Difference of Gaussians (DoG)
 +
* Determinant of Hessian (DoH)
 +
* Maximally stable extremal regions
 +
* PCBR
  
Detecting anything more complex than a point, edge, blob, or corner.  
+
 
 +
 
 +
===Object detection===
 +
 
 +
Tends to refer to detecting anything more complex than a point, edge, blob, or corner.  
 
Regularly by example.
 
Regularly by example.
 +
 +
===Unsorted===
 +
Feature extraction:
 +
* SIFT (Scale-Invariant Feature Transform)
 +
** [http://en.wikipedia.org/wiki/Scale-invariant_feature_transform]
 +
 +
* SURF (Speeded Up Robust Features)
 +
** faster than SIFT, performs similarly
 +
** http://en.wikipedia.org/wiki/SURF
 +
 +
* GLOH (Gradient Location and Orientation Histogram)
 +
** http://en.wikipedia.org/wiki/GLOH
 +
 +
* MSER (Maximally Stable Extremal Regions)
 +
** (primarily blob detection)
 +
** http://en.wikipedia.org/wiki/Maximally_stable_extremal_regions
 +
 +
* LESH (Local Energy based Shape Histogram)
 +
** http://en.wikipedia.org/wiki/LESH
 +
 +
* Scale-space
 +
** http://en.wikipedia.org/wiki/Scale-space
 +
 +
Unsorted
 +
* Hough transform
 +
* Structure tensor
 +
* SPIN, RIFT (but SIFT usually works better{{verify}})

Revision as of 19:12, 8 November 2011

This article/section is a stub — probably a pile of half-sorted notes, is not well-checked so may have incorrect bits. (Feel free to ignore, fix, or tell me)

Feature detection

Where features tend to include:

  • Points -
  • Blobs - smooth areas that won't (necessarily) be detected by point detection. Their approximate centers may also be considered interest points
  • Edges -
    • a relatively one-dimensional feature, though with a direction
  • Corners - Detects things like intersections and ends of sharp lines
    • a relatively two-dimensional kind of feature
  • Ridges

In comparisons between similar images, one should keep in mind that blob-centers can become interest points, gradients can become edges, etc., and that the difference to detectors can be and/or should be fuzzy.

  • Interest point - could be said to group the above and more
    • preferably has a clear definition
    • has a well-defined position
    • preferably quite reproducible, that is, stable under relatively minor image alterations such as scale, rotation, translation, brightness.
    • useful in their direct image context - corners, endpoints, intersections
  • Region of interest


See also:


Edge detection


Interest point / corner detection

  • Harris operator [7]
  • Shi and Tomasi [8]
  • Level curve curvature [9]
  • SUSAN [10]
  • FAST [11]


Blob detection

  • Laplacian of Gaussian (LoG)
  • Difference of Gaussians (DoG)
  • Determinant of Hessian (DoH)
  • Maximally stable extremal regions
  • PCBR


Object detection

Tends to refer to detecting anything more complex than a point, edge, blob, or corner. Regularly by example.

Unsorted

Feature extraction:

  • SIFT (Scale-Invariant Feature Transform)

Unsorted

  • Hough transform
  • Structure tensor
  • SPIN, RIFT (but SIFT usually works better(verify))