Image detection problem using canny edge and Hough Transform

Hi,

I am novice in Matlab and image processing but I am working on detection of iris in a images, These Images are analysed by canny edge for detection of edges and iris and pupil are detected using circular hough transformation.

Most of the images are analysed with proper detection of iris and pupil while some of the images give wrong detection of iris and pupil in an image.

I am not sure where to start looking for the error.

Is there any possibly if I can get some advices where I must start looking for error.

Also please let me know how to upload the Image I am analysing.

[bym]

http://www.mathworks.com/matlabcentral/answers/7924-where-can-i-upload-images-and-files-for-use-on-matlab-answers

Answer 1

[Muhammad shahbaz]

canny edge detection and hough transform are very sensitive with illumination and colour diffusion,better not to use them for iris detection,they normally work for line or sharp edges,increasing threshold value might do something favourable.

Answer 2

[Kaushal]

thanks for the replay Muhammad shahbaz, much appreciated.

Can you please suggest me some other ways of finding edges of the iris and pupil. Also I dont know how to change the threshold value of an image, can you please suggest.

here is the image link of iris http://www.easy-share.com/1917403218/SAD1007iL1siguteleft.bmp and pupil http://www.easy-share.com/1917403178/pupilImage.jpg

The images I am using is are 8 bit gray scales Images taken from Monochromatic Camera

[Sean de Wolski]

your files have been deleted...

Answer 3

[Image Analyst]

Try Libor Masek's iris recognition package: http://www.csse.uwa.edu.au/~pk/studentprojects/libor/

[Kaushal]

Thanks Analyst.

I am also working on it, but there is problem in segmentation process.

As Images are not segmented properly.

It happens the code written by Libor Masek is good one but it greatly depends on threshold value. I am also trying to improve his work so that it could be able t detect iris, in with any thresholds. Also I feel some how the canny edge is not detecting properly, wishing to improve it as well.

I am thin of using shen castan algorithm for edge detection.

Much appreciate if you can comment on these thoughts and guide me in correct direction.

[Kaushik Acharya]

By thresholding do you mean the radius range of pupil, iris which Libor has kept different for different databases in segmentiris.m?