| ProPortion (version 2.2.0) is software that performs multiple-category segmentation in colour images. Segmentation is a term that is used when an image is divided into constituent regions. The lumber image below has been segmented into two regions: (i) the wood and (ii) highlighted in red are the knots and splits. |
 |
Also - take a look at the case study to see ProPortion in action. And download it here (free 60 day trial).
Installation
ProPortion is stand-alone software. It will run on your work computer without IT support and configuration.
Download the self-extracting ZIP file to your computer's hard drive. We will assume that you used the default folder,
C:\ProSensus\ProPortion2.2.0\
to extract the files. But you may, of course, chose any other location.
To start using ProPortion, change to the directory where the software was installed and double-click on
ProPortion.exe
System requirements
At this time, ProPortion is only available for Windows operating systems. If you can run Windows 2000 or XP, then you can run ProPortion.
To uninstall ProPortion
Delete the folder where the ProPortion files were unzipped to, eg C:\ProSensus\ProPortion2.2.0\
Need the manual ?
All this information is reproduced in the PDF manual.
How does it work ?
About digital images
Before describing how ProPortion works it is important to understand how digital images are represented on your computer. Colour image files are just a series of numbers. These numbers are arranged in three layers, red, green, and blue, which is why they are also called RGB images. Each of these layers provide information about that colour level when it was recorded by the camera. The figure below shows illustrates this.
A magnified part of the lumber image, showing the numbers that represent the R G B layers
From the human eye to a digital image
The human eye experiences a near infinite number of colour levels, but a digital image has to make some compromise: each layer can only have 256 unique levels. Each level in the RGB layers are labeled as 0, 1, 2, ... , 255. This is still a very large number of colours: a total of 2563, or 16 777 216 unique colours are possible. For reference,
| • a pure black colour has | (0,0,0) = (R,G,B) |
| • pure white | (255,255,255), |
| • red | (255,0,0), |
| • green | (0,255,0), |
| • blue | (0,0,255), |
| • yellow | (255,255,0), and |
| • purple | (255,0,255). |
Colour segmentation
The purpose of colour image segmentation is to find a range of colours in the image and isolate them from the rest of the image. For example, if you know that a certain defect has a red colour, then ProPortion can find this red colour. But, this red colour is not just 1 of the 16 777 216 unique colours, because there are many shades of red.
So, how does ProPortion work ?
There are step-by-step instructions in this case study. But here is the bottom line:
- You tell ProPortion which range of colours to find by selecting a part of the image.
- ProPortion uses this range of colours to build a model.
- This model is applied to the entire image to isolate all similar colours.
- There is a margin factor setting that controls how strictly the model is applied to the image.
- The model is automatically expanded to handle multiple regions.
What type of models are used? ProPortion uses PLS models, projection to latent structures, which are extremely effective for dealing with multivariate data.
