Monday, December 8, 2014

Enhancement: Subtract Background

In ImageJ, non uniform illumination may be corrected with the function Process > Subtract Background...

The Subtract Background function is a filter called a 'Rolling Ball' filter and is equivalent to a 'Top Hat' filter  [Wikipedia]. There is an explanation of its behavior in ImageJ User Guide [Link].

Fig.1: Dialog window of the Process > Subtract Background... function.

1. Choosing the good radius

Like all the filters, the main parameter is the convolution mask radius (by default, 50). How to choose the best radius?
The principle of this filter is to roll a circular probe along the bottom of the profile (Fig.2) to extract the background image. Then, this background is subtracted to the input image.

Fig.2: Principle of the 'Rolling Ball' filter.

If you choose a too small radius, the probe is "entering" in the peaks corresponding to the objects of interest and they will be included in the background image.
If your radius is too large, only the main features are taken into account and local variations won't be included.

2. Other options

Several other options may be set up depending of the type of image to enhance.

The first question to ask is if the background is darker or lighter than the objects of interest to (un)check the 'Light Background' checkbox. If you choose the wrong option, square artifacts appear as shown in Fig. 3.

Fig.3: Square artifacts due to an inverted background.

If the filter doesn't work as expected, it can be interesting to look at the background image. That's what is doing the 'Create background (don't subtract)' option.

3. Other crazybiocomputing posts

Further readings are available in ...
  • Image Enhancement Series  [Link]
  • Image Processing TOC [Link]

No comments:

Post a Comment