Unable to identify the iris region in human eye images

Question

Using OpenCV to identify the iris region + pupil region (outer grey area + inner black circle) as seen in this image

Tried the following approaches, but unable to extract the iris region 100%.

Approach 1 Iris area detection using detection of color code of the pixels in the image

import cv2
from PIL import Image
#import cv2.cv as cv

img = cv2.imread('i1.jpg')
im = Image.open('i1.jpg')
pix = im.load()
#cv2.imshow('detected Edge',img)

height, width = img.shape[:2]

print height,width
height=height-1
width=width-1
count=0
print pix[width,height]
print pix[0,0]
for eh in range(height):
    for ew in range(width):
        r,g,b=pix[ew,eh]
        if r<=30 and g<=30 and b<=30:
            print eh,ew
            cv2.circle(img,(ew,eh),1,(0,255,0),1)
print height,width
cv2.imshow('detected Edge',img)
cv2.waitKey(0)
cv2.destroyAllWindows()

Click here to view output of the above code.

Approach 2 Iris area detection using Hough Circles method

import cv2
#import cv2.cv as cv
img1 = cv2.imread('i.jpg')
img = cv2.imread('i.jpg',0)
ret, thresh = cv2.threshold(img, 50, 255, cv2.THRESH_BINARY)

edges = cv2.Canny(thresh, 100, 200)
#cv2.imshow('detected ',edges)


cimg=cv2.cvtColor(img, cv2.COLOR_GRAY2BGR)

circles = cv2.HoughCircles(edges, cv2.HOUGH_GRADIENT, 1, 10000, param1 = 50, param2 = 30, minRadius = 0, maxRadius = 0)
#circles = cv2.HoughCircles(img,cv2.HOUGH_GRADIENT,1,20,
#                            param1=50,param2=30,minRadius=0,maxRadius=0)
print circles
for i in circles[0,:]:
    i[2]=i[2]+4
    cv2.circle(img1,(i[0],i[1]),i[2],(0,255,0),1)

#Code to close Window
cv2.imshow('detected Edge',img1)
cv2.waitKey(0)
cv2.destroyAllWindows()
for i in range (1,5):
    cv2.waitKey(1)

Click here to view the output of the code.

Kindly guide us how we can automatically extract the circular black area in human eye pictures.

that is not the cornea. that's the pupil... are such things not teached in schools anymore? :( and what is wrong with the hough output image? what do you want to achieve if this result is not good enough and what is good enough? — Piglet
To be precise we want to identify the iris region(outer circle grey area) and the pupil region (inner black circle) as seen in this image www3.sciencedump.com/sites/www.sciencedump.com/files/plaatjes/…. Hough output images are not consistent. In some eye images its not precisely identifying the iris region. — Vikram
you can't expect good results when you don't set min and max radius for hough transform... — Piglet

andrew andrew · Accepted Answer · 2016-04-26T20:32:09

Using approach 2, could you start at the center of the pupil and then travel outwards staying in the same row (travel left or right of the pupil center) until you hit the sclera of the eye. Use this as the radius for the circle containing the iris.

radius_iris = abs(first_column_of_sclera - column_of_pupil_center)
#this is the yellow line in the attatched image

To Find the Sclera: take a small pixel region like a 3x3 block (or similar, this is the green box in the image) and check for two criteria

The variance of the r,g,b channels is small. White (or gray shades) have R=G=B so that means white would have low variance
You also need to check that the rgb value is above some threshold. Someone with grey or black eyes will meet criteria 1, but unless the pixels are very light (near white) we haven't reached the sclera

Create an iris mask by creating a circle centered at the pupil with radius_iris if you want, you can also use the pupil mask to extract ONLY THE iris

Unable to identify the iris region in human eye images

3 Answers