This site is devoted to mathematics and its applications. Created and run by Peter Saveliev.

Images appropriate for analysis

From Intelligent Perception

Jump to: navigation, search

The analysis method used by Pixcavator has certain limitations. The two principal ones are descried below.

1 Dimensionality

We do not extract 3D information from 2D images.

This means that the analysis is meaningful only when the image can be interpreted as 2-dimensional.

Examples of appropriate images are:

Dros.JPG Fp0.JPG Petri.jpg

In inappropriate images the third dimension is essential. They may contain:

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

Mugshots may serve as an example of appropriate images of 3D objects because they are always taken under same angle with similar lighting. For the same reason, applications in machine vision for industrial inspection are also appropriate. Detecting a change in a 3D scene (appearance of a person in a security camera) is also feasible.

Images of 3D objects may contain 2D items: text, scratches, other imperfections of the photograph or the lenses, noise, etc.


2 Proximity

As it often happens, images are shrunk from the original. In this case, a smaller image is acquired via averaging which leads to blurring. There are two possible consequences.

First, it may be impossible to separate objects such as particles. Even though two particles may seem distinct to you, if you zoom in, you might see a thin bridge between them. As a result the software thinks this is the same particle.

We do not separate objects.

Also,

We do not group objects.

Pixcavator does not treat two objects as one, no matter how close they are. For example, these are inappropriate images:

  • a thin scratch on the photo will cut an object in two,
  • person's shirt and pants (of different color) will cut him in half,
  • the background of an object is uneven and has colors similar to the one of the object.

The simplest way to group two adjacent objects into one or separate two objects connected by a thin "bridge" is via dilation and erosion. Pixcavator has these commands in Tools tab.

At this time these tools exist as a way to modify the image prior to analysis. These "morphological operations" aren't integrated in Pixcavator yet. Nor are watershed algorithm and similar.