Equalize images

[DocEye]

Hi,

prob. another silly question.
I have an image in my sequence (eyes from different angles) which is darker then the other ones. The nose shadows my illumination. Is there a way to analyse the other images and then adjust the parameters of the dark image according to that. Would make them more compareable and keep the edge detection constant.

thanks for your answers

Fabian

[RManion]

Hello again Fabian!

Yes, preprocessing can often be a very important step of an image processing algorithm. Regarding the particular problem you are referring to, it may be possible to use an IMAQ Equalize to create a similar looking image each time.

The IMAQ Equalize function is located on the Processing subpalette of the Image Processing palette.

Give it a try and let me know how it goes. Hopefully the function can at least maintain a similar lighting scheme.

Robert
Applications Engineering
National Instruments

[DocEye]

Thanks Robert,

unfortunately the imaq equalize function doesnt work for that problem. I would like to analyse the "good" image and apply the histogramm to the "bad" image. I hope that this would improve the colorpatternmatching function and make the edge detection more constant. I attached 2 images, maybe you could have a look.

thanks for your help

[RManion]

Hi Fabian,

Unfortunately, if you start with a bad image, you will normally end up with bad results. From the images you have included, I would suggest some kind of lighting to properly illuminate the eye. You might also try to keep the face at a particular angle. It doesn't look like you will be able to do a great deal to improve the quality of the image, but perhaps your edge detection will work better on a particular color plane.

Good luck!

Robert

[DocEye]

Thanks again Robert,

I know that the images are not in a perfect quality but the lighting is a bit difficult since a brighter light dazzles the patient, also I sent you one of my first pictures, the focus is a bit better now. But I still have the problem with that one "bad" image.
My edge detection works already with the colorplain which has the best contrast for finding the edges.
My problem is rather how to improve the bad image to find the pattern learned in the good image. I am thinking about somehow analyse the histogram and than apply BCG transformation to the "bad" image to equalize the histograms....maybe.
I can improve the bad image manually so that the pattern matching works fairly food. I am just looking for a way to improve the bad image during run time automatically before being analysed.

Thanks again for your help and patience

Fabian

p.s. could you maybe have a brief look at this post and try to give me a suggestion.
http://forums.ni.com/ni/board/message?board.id=200&message.id=6238

[Nicolas]

Fabian, here is a link to an histogram matching example posted a few weeks ago by Dinesh:
http://forums.ni.com/ni/board/message?board.id=200&message.id=5513&requireLogin=False

Nicolas