Enhancement: Non uniform illumination

In the Image Enhancement series, the non uniform illumination is the second main defect that you must correct in your images.



[TODO]
Fig. 1: Test Image

The test image of Fig.1 has a problem of non uniform illumination as shown by the horizontal profile of Fig.2.

[TODO]
Fig. 1: Horizontal plot profile of the background of image in Fig.1.

Several techniques are available to try to flatten the baseline of the profile and can be sorted in two categories:

1. With a reference image

If you are using a scientific device like a microscope (optical or electron), all the environmental parameters are well known (no clouds, fog, sun,etc.). In this case, you can record a blank image at the beginning of your session of images collection. A blank image or reference image is an image without samples. Thus, if there is a problem of illumination, this image contains the defect.

Note: With this technique, we assume that the defect is constant during the session.

All the images collected during the session have exactly the same defect as in your blank image. To fix the defect, just subtract the blank image to your sample images with the Process > Image Calculator.

2. Without a reference image

Unfortunately, the reference image is outdated or you didn't backup this image or you didn't know where this image is located in the hard disk. So, the result is: You have no reference image!

All the techniques available tried to model the background image and then to subtract it to the images.

• Use filters like the 'Rolling Ball' or the 'Top Hat' filters.
• Model the background with a polynomial function.
• Use of Mathematical Morphology method.
• Removal of low frequencies in frequency - Fourier - space.

In ImageJ, only the 'Rolling Ball' filter is implemented to subtract the background.

<<  TOC : PreviousNext: Rolling Ball filter >>

3. Other crazybiocomputing posts

Further readings are available in ...

• Image Enhancement Series  [Link]
• Image Processing TOC [Link]