uneven illumination in pictures

[watertreader]

uneven illumination correction

Hi,

How do we generally deal with uneven illumination in pictures? ie one corner of the picture/image is brighter than the rest

 

[auto_mitch]

Hi watertreader.
The uneven illumination is one of the most serious problems in image processing. I will describe in the following lines a method for grayscale images that i have already applied to my project.
Image (f(x,y)) is a function which is consist of two components1) the amount of source light incident on the scene being viewed and (2) the amount of light reflected by the objects in the scene. Appropriately they are called the illumination (i(x,y)) and reflectance (r(x,y)) components.
The functions i(x,y) and r(x,y) combine as a product to form f(x,y):
f(x,y)=i(x,y)*r(x,y)
Our goal is to have at the end only the reflectance component, so initially must be calculated the illumination function. At the frequency domain the illumination is the very low frequencies and the reflectance is the high frequencies.
The initial image is filtered by a very low pass filter (F(0,0)) in frequency domain. The result is an image which shows the estimated illumination.
As you can see at the above equation the two components are combined by non linear method (multiplication).
If you divide the initial image with the estimated illumination you will get an image with illumination correction. The result of the division is the reflectance component.

Regards
auto_mitch

 

[watertreader]

Would this method be similar to another kind of method which people refer to as flat fielding? What's the difference?

 

[auto_mitch]

Would this method be similar to another kind of method which people refer to as flat fielding? What's the difference?

I think that it is the same method. I have already try two more methods. Try to find literature for "Homomorphic Filtering".
Another method is to calculatate the uneven illumination gradient with morphological operations. If the image objects have all the same local contrast (if they are either all darker or brighter than the background), top-hat transforms can be used for mitigating illumination gardients. A top-hat with a large structural elementacts acts a high pass filter. White top-hats are used for dark backgrounds and black top-hats for bright backgrounds. At the white tophat method is used a morphological opening method with a large structural element and the result is subtracted from the original image. Finally you have an image with homogeneous illumination. At the black top-hat is used a morphological closing method to calculate the illumination gradient.

Regards
auto_mitch