This page is a part of CVprimer.com, a wiki devoted to computer vision. It focuses on low level computer vision, digital image analysis, and applications. It is designed as an online textbook but the exposition is informal. It geared towards software developers, especially beginners, and CS students. The wiki contains mathematics, algorithms, code examples, source code, compiled software, and some discussion. If you have any questions or suggestions, please contact me directly.

Analysis of sample images

From Computer Vision Primer

Jump to: navigation, search

During processing, the areas, perimeters, and contrasts of the objects are also computed. These parameters are used filter objects. The user indicates ranges of parameters of the features he considers to be irrelevant or noise.

Coins.
Enlarge
Coins.
 The coins are captured.
Enlarge
The coins are captured.

The settings may be chosen based on a priori knowledge about the image. For example the image in Fig. 21 is 300×246. To capture the coins and ignore the noise and the small features depicted on the coins, one sets the lower limit for the area at 1000 pixels.









Coins blurred.
Enlarge
Coins blurred.
 The coins are captured.
Enlarge
The coins are captured.













Coins with Gaussian noise.
Enlarge
Coins with Gaussian noise.
 The coins are captured.
Enlarge
The coins are captured.













Enlarge
Enlarge

The analysis is not significantly affected by rotations. The output for the original 640×480 fingerprint in Fig. 24 is 3121 dark and 1635 light objects. For the rotated version, it is 2969-1617. By limiting the analysis to objects with area above 50 pixels, the results are improved to 265-125 and 259-124, respectively.






Stretching the image does not affect the count of objects. Shrinking makes objects merge. If the goal, however, is to count and analyze larger features, limited shrinking of the image does not affect the outcome. The count is also stable under noise and blurring.

The method works best with images that represent something 2-dimensional. It fails when the third dimension is essential. It is the case when the images contain:

  • occluded objects,
  • transparent objects,
  • objects well lit on one side and dark on the other,
  • X-rays.

Another limitation of the method is that clustered cells or other objects can’t be separated unless there is variation of intensity.





Other case studies

Personal tools