XOR the iris-mask.

thank you for your response Mr.Lambert Zijp...But,XOR operation is suitable for circular mask with fixed radius iris region..is there any method to create the circular masks automatically according to the radius of the iris region..in my work large number of images were used..The circle detection method based on circular hough transform detects the boundaries of the iris regions automatically..however,the cropping section need human interaction to choose the circular masks..is there any approach to mask the region out side the circle detected by circular hough transform..

Maybe I don't understand the problem again...

For each image you detect the iris using the Hough transform. So you have an X,Y center and a radius. Use them!

Create a second image of equal size, loop over the X and Y, if an XY pair is inside the circle, set it to 1 or 0 depending on whether you want to mask the iris or you want to mask the non-iris.

-- or --

Create a second image of equal size, draw the circle, floodfill the image using the upper-left pixel as a seed. Maybe XOR the result depending on whether you want to mask the iris or you want to mask the non-iris.

i am using UBIRIS database..the database contain images acquired at various distances ..the iris diameter is varies according to the image acquisition distance..for these images,the iris diameter varies from 90 pixels to 220 pixels..the Circular Hough transform based approach is suitable to detect the iris circles in all the images(even though their radius is unique)..after detect the iris circle,i want to crop the iris region,masking of non iris region is also enough..regarding this, a circular mask is created..every time i create a separate mask,because iris diameter is varies according to the image acquisition distance..it is difficult to create mask every time(database contains more than 11000 images)..is there any technique or matlab code is available to create these masks automatically according to the variable iris diameter..

The method I described above shouldn't take more than a second. Your database will then be done in three hours. (You can optimize it: determine the bounding box, -radius to +radius, only compute one quadrant, use symmetry).

Maybe transforming a high-resolution prepared circular mask with a transformation matrix containing a scaling and two translations, using each image as a resampling grid is faster... You need to try.

thank you..i will try that

"roicirclecrop"   is function in matlab,   it allows to crop a circular region

Thank you Lamis..but,roicircle crop need a human interaction   

sir whether u find ur answer ? plz share ur answer 2 me bcz i have same prblm that's u had? so plz help me....

@vikas Gupta

step(i) Detect the circle using CHT

step(ii)set the outside pixel values into zero.(here you have to use the parameter radius of the circle).

you can also create the mask using ROI selection ...