How to get glare/reflection free image from multiple images


I am working on concept where i have 6 Images which are captured under 6 different illumination condition. Different illumination causes reflection in different places in a captured images. So i have use those images and get one reflection free image without losing/altering any information of image.i did lot of experiments on the collected 6 images but nothing worked…any suggestion it could be helpful for me…Thank you

Hi Usha,

I do not know what kind of images you capture and in what circumstances, but it is possible to cover the illumination with polarization foil and use a camera lens with polarizing filter? If the polarized light is the only light source, almost all the reflection will disappear.

Hello Usha -

This sounds like a very interesting image-processing challenge,
not necessarily easy, but possibly doable. (I’m not aware of
anything built in to Fiji / ImageJ that would do this out of the box.)

It would be helpful if you could post your six images. That
might help forum participants see if they can come up with
any suggestions.

A very important question is whether most (say four out of the
six) images are free of glare in any given region of the image.
This would make the processing easier, in that you could use
consistency between most of the images – say some kind of
majority vote – to reconstruct the single “clean” image.

But if regions of the image often have glare in most of the images
it becomes harder because you would need to develop some
technique to distinguish glare from non-glare.

Thanks, mm

1 Like

Hi, Thank you for your quick response.

As i said previously, all images contain glare (at same and different region).

we used vote based reconstruction but it won’t worked.

Now we r looking to design an approach to find glare and non-glare area…

if any suggestion on this, please let me know.

Thank you

Hi Peter,

Thank you for your quick response.

The main challenging thing is we have to use only aquired images(without hardware change) to get a glare free image.

Hello Usha -

It’s hard to make any concrete suggestions without some example
images to consider. The task could be hard, and a potentially
successful approach could depend on the details of your images.

You mention finding the “glare and non-glare area.” This seems
to fall into the category of “image segmentation.” Whether the
following might work will depend on the details of your images
and whether you have enough training data.

There is a segmentation plugin (that I have never used) that
ships with Fiji / Image:

Plugins > Segmentation > Trainable Weka Segmentation

Here is some documentation:

Trainable Weka Segmentation

The basic idea is that you have some images that you have
manually segmented into “glare” and “non-glare” areas.
From these manually “annotated” images the plugin “learns”
(gets “trained”) to segment images that you haven’t processed

(Again, I have no idea how well the Trainable Weka Segmentation
plugin works, and whether it would be appropriate for your problem.)

(If you can get the segmentation to work, I believe your intent
would be to use the non-glare sections of your six images to
construct your composite glare-free full image, and such a
scheme could certainly make sense.)

Thanks, mm