Object Extraction

In JMicroVision components can be extracted as objects (e.g. particles) with the Object Extraction tool or as backgrounds (e.g. matrix) with the Background tool. The main differences are that the Object Extraction tool transforms objects into vector drawings and computes more than forty descriptors on each object. However, it is recommended to use the Background tool when only the entire surface of a component needs to be calculated and it is not necessary to partition the background in multiple regions.

The Image Factory tool offers some image preprocessing functions and different segmentation methods for more advanced use (see tutorials on the website).

 Note: Point counting is an alternative method to quantify image components, especially when the image is too complex to use the Object Extraction or Background tools.

Extraction

The methods in the list enable creating objects (vector graphics) in different ways. Once created, they can be classified (see Classification) and their descriptors can be displayed in the Data Viewer.

Start: Open the dialog of the selected method.

Class Editor: Make your own list of classes.

Intensity threshold

The segmentation by threshold is the method commonly used to extract objects. The user selects a range of values in a histogram. This range must match as well as possible the pixel intensities of the objects to extract (see Histogram Thresholding). In addition, other parameters can be applied:

When the Extraction config item is selected, a summary of all the parameters values is displayed. This configuration can be saved with the Save as Template button. Afterwards, the template will appear in the combo box (even after closing JMicroVision) and when it is selected, it reapplies the saved configuration.

 Note: Templates can be used by the Object Extraction or the Background tool, but they can only be created or modified by the Intensity threshold dialog.

 

Convert 2 dimensions drawings

This method allows converting 2D drawings into objects. When the segmentation methods are not able to extract objects correctly, you can manually create an outline around objects with 2D Measurement tool and then convert them to obtain more descriptors.

 

Magic Wand

See Magic Wand.

 

Object Separation

This experimental module is similar to Intensity threshold, but has extra functions to separate connected objects.

 Note: The separation process requires to handle the whole image in memory, thus it can be applied on very high-definition images only if the computer has enough memory.

Separation: Choose between Variable (undefined shapes) or Custom Shape, if the objects to separate have common shape characteristics. The shape variance is defined by a set of 2D Measurement drawings or a class of objects, which are specified when selecting Custom Shape.

Shape Size: Define the Min size and the Max size of the objects.

 Note: When the Variable Shape is chosen in Separation, the displayed shape is an ellipse that actually represents undefined shapes. For Custom Shape, the displayed shape is a vector drawing, which represents the mean of shape parameters of the drawings.

Shape Tolerance: Define the shape and the eccentricity (elongation) tolerance of the objects to separate.

 Note: For Variable Shape, the sliders are locked to 100% (because the shape is not defined). For Custom Shape, the sliders initial values are proportional to the shape and eccentricity variance of the set of vector drawings.

The other commands are identical to Intensity threshold.

 Note: Templates can be saved and reapplied later (in Separation config).

Classification

Allows classifying the objects, see Classification.

Objects

Color: Color all the objects according to a parameter, see Coloring Objects.

Map: Color points (points represent the center of gravity of the objects) according to a parameter.

Modify: Modify the object shape by one of these operations:

  • Smooth (mean): Smooth the contour line of the shape and the holes by a mean filter. The filter size is 2*Disc radius + 1 (in pixels).
  • Smooth (Gaussian): Smooth the contour line of the shape and the holes by a Gaussian filter. The filter size is 2*Disc radius + 1 (in pixels). A Gaussian filter preserves more the contour shape than the mean filter.
  • Polygonal approximation: Transform the outline of the shape by a polyline. If the shape contains holes, they are removed. The Disc radius represents the approximation value of the shape outline. Increasing this value will reduce the accuracy of the shape.
  • Convex hull: Transform the outline of the shape by a convex outline and if the shape contains holes, they are filled.

Object Extraction preferences

Compute Additional Parameters: see Object Descriptors.

Fill drawing: When the button is unchecked, only the outline of objects is displayed.

Only the checked parameters will appear in the Selected Object panel and in the Data Viewer.

 Note: To display the preferences, open the Project Preferences window , select Object Extraction.

 

See also:

Object Descriptors, Feret Diameters, Context, Texture