Thread Subject: New Iris Recognition Results

Hello,

Recently, I proposed a new iris recognition system based on Circular Fuzzy Iris Segmentation and Gabor Analytic Iris Texture Binary Encoder. The following figures illustrate the latest results obtained by me:

Circular Fuzzy Iris Segmentation:

http://www.limbistraine.com/ro/software/popescu-bodorin/img/circular-fast-fuzzy-iris-segmentation-popescu-bodorin-nicolaie.jpg

Gabor Analytic Iris Texture Binary Encoder:

http://www.limbistraine.com/ro/software/popescu-bodorin/img/bath-university-iris-database-two-iris-codes-gabor-analytic-iris-texture-binary-encoder-popescu-bodorin-nicolaie.gif

The latest iris recognition results obtained using eye images from Bath University Iris Database:

http://www.limbistraine.com/ro/software/popescu-bodorin/img/bath-university-iris-database-a-new-iris-recognition-system-based-on-gabor-analytic-iris-texture-binary-encoder-popescu-bodorin-nicolaie.gif

Your ideas, comments, criticism, opinion, collaboration are welcomed ( bodorin [#] ieee . org ).

Best regards,
Nicolaie Popescu-Bodorin
Teaching Assistant, Math & Computer Science Department,
Spiru Haret University, Bucharest, Romania

Forbidden
You don't have permission to access /ro/software/popescu-bodorin/img/
circular-fast-fuzzy-iris-segmentation-popescu-bodorin-nicolaie.jpg on
this server.

Additionally, a 404 Not Found error was encountered while trying to
use an ErrorDocument to handle the request.
--------------------------------------------------------------------------------

Apache Server at www.limbistraine.com Port 80

Hello ImageAnalyst,

Thank you very much.
Here is a web page containing those figures:

http://www.limbistraine.com/ro/software/popescu-bodorin/fuzzy-iris-segmentation-gabor-analyticiris-texture-binary-encoder-results-for-bath-university-database.htm

Best regards,
Bodorin

Hello ImageAnalyst,

I solved the issue. A misplaced HotLink Protection instruction on my site was responsible for that behavior. Thank you again.

Best regards,
Bodorin

On May 17, 5:37 pm, "Nicolaie Popescu-Bodorin"
<freeseosub...@yah00.com> wrote:
> Hello ImageAnalyst,
>
> Thank you very much.
> Here is a web page containing those figures:
>
> http://www.limbistraine.com/ro/software/popescu-bodorin/fuzzy-iris-se...
>
> Best regards,
> Bodorin

-------------------------------------------------------
This one works. It did a nice job of finding the iris in the two
photos. The irises look rather unusual. I've never seen one with
such a spotted/mottled appearance. Do you have examples with more
normal-looking irises, with radial striations? It would be good to
see more examples so one can get an idea of how robust your method is,
as far as being able to handle a variety of different looking
subjects.

I have no idea what the binary iris code images represent. Do you
have a 3 or 4 line description of how they were derived, what they
represent, how to interpret them, or how to compare them?

Did you share the MATLAB code somewhere? I didn't see it.
Regards,
ImageAnalyst

Hello ImageAnalyst,

I'm very glad to see that you are so familiar with the iris recognition problem.
The reason for which the irises look that way is the fact that the figure contains subsampled versions of the original images (each original image is 960x1280 and is impossible to fit two of this kind in the same figure).
If you whish to see the original images, you may apply for the free version of Bath University Iris Database (BUID) containing 1000 such normal-looking irises examples.

The third figure shows the robustness in terms of recognition scores (for all-to-all comparisons between the irises taken from all eye images within BUID).

You cuoldnt know now what the binary iris code images represent because Gabor Analytic Iris Texture Binary Encoder is a new method for encoding the iris texture. It will be published soon. The idea is to extract the phase of carrier wave of iris texture at certain scale. Iris similarity is measured using Hamming distance.

The reason for posting the current thread in the forum is finding collaborators on these subjects (joint works, code/data/knowledge exchange, results replication / comparison, research grants, etc). Still, the code will not be publicly available.

Do you have your own iris recognition results? You seems to have a lot of experience in this field. Have you worked with BUID allready? Do you work with other iris datase? If yes, do you agree with exchanging iris segment collections?

Regards,
Bodorin
 
 <imageanalyst@mailinator.com> wrote in message <bae7c21e-b963-4220-b931-1a54363b849f@l28g2000vba.googlegroups.com>...
 
> -------------------------------------------------------
> This one works. It did a nice job of finding the iris in the two
> photos. The irises look rather unusual. I've never seen one with
> such a spotted/mottled appearance. Do you have examples with more
> normal-looking irises, with radial striations? It would be good to
> see more examples so one can get an idea of how robust your method is,
> as far as being able to handle a variety of different looking
> subjects.
>
> I have no idea what the binary iris code images represent. Do you
> have a 3 or 4 line description of how they were derived, what they
> represent, how to interpret them, or how to compare them?
>
> Did you share the MATLAB code somewhere? I didn't see it.
> Regards,
> ImageAnalyst

On May 17, 7:50 pm, "Nicolaie Popescu-Bodorin"
<freeseosub...@yah00.com> wrote:
> Hello ImageAnalyst,
>
> I'm very glad to see that you are so familiar with the iris recognition p=
roblem.
> The reason for which the irises look that way is the fact that the figure=
 contains subsampled versions of the original images (each original image i=
s 960x1280 and is impossible to fit two of this kind in the same figure).
> If you whish to see the original images, you may apply for the free versi=
on of Bath University Iris Database (BUID) containing 1000 such normal-look=
ing irises examples.
>
> The third figure shows the robustness in terms of recognition scores (for=
 all-to-all comparisons between the irises taken from all eye images within=
 BUID).
>
> You cuoldnt know now what the binary iris code images represent because G=
abor Analytic Iris Texture Binary Encoder is a new method for encoding the =
iris texture. It will be published soon. The idea is to extract the phase o=
f carrier wave of iris texture at certain scale. Iris similarity is measure=
d using Hamming distance.
>
> The reason for posting the current thread in the forum is finding collabo=
rators on these subjects (joint works, code/data/knowledge exchange, result=
s replication / comparison, research grants, etc). Still, the code will not=
 be publicly available.  
>
> Do you have your own iris recognition results? You seems to have a lot of=
 experience in this field. Have you worked with BUID allready? Do you work =
with other iris datase? If yes, do you agree with exchanging iris segment c=
ollections?
>
> Regards,
> Bodorin
>
---------------------------------------------------------------------------=
-----------------------
Borodin:
I do have lots of experience in the image analysis field, having
worked on hundreds if not thousands of industrial image analysis
projects over the past several decades. The last time I worked on
eyes though was about 10 years ago when I had to find the pupils to do
some work on the face, and before that in the mid-80's for a couple of
years when I was developing a laser scanning ophthalmoscope. But we
looked at the retina, not the iris. I haven't had a project that
worked directly on the iris. It seems more like a security/
authentication application, and I'm not really in that field. I'm
more in industrial scientific R&D. I was just curious because I'm a
voracious learner. Good luck with your work.
ImageAnalyst

On May 17, 7:50 pm, "Nicolaie Popescu-Bodorin"
<freeseosub...@yah00.com> wrote:
> Hello ImageAnalyst,
[snip]
 Iris similarity is measured using Hamming distance.
>
[snip]
> Regards,
> Bodorin
>
---------------------------------------------------------------------------=
-------------------
Can you expain why Hamming distance is used to compare the "test"
image to the "template" iris image? What advantages does that have
over other image comparison methods such as PSNR, cross correlation,
or similar kinds of things?

Hello ImageAnalyst,

Hamming Distance vs. PSNR:

Let consider PSNR first. It is usualy used as a measure of image distortion.
Typically, PSNR is used to evaluate the quality of a reconstructed image when compared to the original image.
More general is the case when one of the images is considered a noisy approximation of the other but the same object is represented in both images.
This assumption is not true for iris images.
Here the most important difference between two iris images comes from the fact that, even in the most favorable practical scenario, two iris images represent two distinct hypostases of the iris i.e. two distinct objects.
Comparing two iris images using PSNR doesn't make much sense if accurate motion law IrisFeatures-PupilDilatation is not available. But comparing two iris binary templates using PSNR or Hamming Distance is the same thing.
There are a lot of papers investigating how image compression (i.e PSNR) modifies the recognition accuracy.

Hamming Distance vs. cross-correlation:

I still study this subject. I belive that interesting results will arise from this area (at least in terms of testing the robustness of any recognition method).
As an example, the following figure shows some recognition results obtained durring the previous month for Bath University Iris Database when I used a custom modified Hamming-correlation distance:

http://www.limbistraine.com/ro/software/popescu-bodorin/img/bath-university-iris-database--recognition-results--custom-modified-hamming-distance--popescu-bodorin-nicolaie.gif

But, despite the above results, bigger computational effort is a strong reason for avoiding the use of correlation in real-time applications.

Regards,
Bodorin

______________________
ImageAnalyst wrote:
> Can you expain why Hamming distance is used to compare the "test"
> image to the "template" iris image? What advantages does that have
> over other image comparison methods such as PSNR, cross correlation,
> or similar kinds of things?

Hello Image Analyst,
Can you suggest me please some similarity measures other than hamming, correlation and psnr?
Regards,
Bodorin

Circular Fuzzy Iris Segmentation Demo Program available for download:

http://fmi.spiruharet.ro/bodorin/arch/cffis.zip

Your ideas, comments, criticism, opinion and collaboration are welcomed
( bodorin [#] ieee . org ).

Best regards,
Nicolaie Popescu-Bodorin
Teaching Assistant, Math & Computer Science Department,
Spiru Haret University, Bucharest, Romania

Hi all,
I m working on Iris Recognition technique......
and i m able to achieve a reasonable accuracy...
now, our teacher has told us to try to implement SEGMENTATION TECHNIQUE....
I have read many pprs on the same,
but i will be grateful if some one pllzzz provide me with a general way of going about it...so dat it cn be implemented in matlab 7.....
THNK YOU.....
 

On Aug 27, 7:55 am, "Abhi " <abhigra...@gmail.com> wrote:
> Hi all,
> I m working on Iris Recognition technique......
> and i m able to achieve a reasonable accuracy...
> now, our teacher has told us to try to implement SEGMENTATION TECHNIQUE....
> I have read many pprs on the same,
> but i will be grateful if some one pllzzz provide me with a general way of going about it...so dat it cn be implemented in matlab 7.....
> THNK YOU.....

-----------------------------------------------------------------------
Segmentation is basically just finding the object your interested in
among all the other objects, clutter, and background in your image.
Yes, I bet you will find lots of papers on this.
You're probably already doing it since you say you're getting the iris
with reasonable accuracy.
Perhaps you just didn't realize you were doing it because of the name.

The general way is to
1. Find something about your object that is different from all the
other stuff in the scene.
2. Use that to create an image that you can then threshold to get a
binary image.
3. Do connected components labeling on that binary image of your
object.
4. Measure the objects, for example with regionprops().
It's possible to do this with MATLAB but you'll need to replace your
defective keyboard or you'll experience hell with all the syntax
errors you're going to get. Either that or get a computer because
MATLAB doesn't run on mobile phones as far as I know.

The images from above replica(s) have been moved to following locations:
http://fmi.spiruharet.ro/bodorin/myimg/circular-fast-fuzzy-iris-segmentation-popescu-bodorin-nicolaie.jpg

http://fmi.spiruharet.ro/bodorin/myimg/bath-university-iris-database-two-iris-codes-gabor-analytic-iris-texture-binary-encoder-popescu-bodorin-nicolaie.gif

http://fmi.spiruharet.ro/bodorin/myimg/bath-university-iris-database-a-new-iris-recognition-system-based-on-gabor-analytic-iris-texture-binary-encoder-popescu-bodorin-nicolaie.gif

Regards,
Nicolaie Popescu-Bodorin
Teaching Assistant, Math & Computer Science Department,
Spiru Haret University, Bucharest, Romania