Locating ROI in Iris Using Randomized Hough Transform

Abstract

Iris segmentation is applied to the human eye to extracting the region of interest (ROI) and is used to recognize the human uniquely and efficiently. In this, we have first detected and removed the reflection mask using adaptive thresholding; then to detect edges, the Gaussian filter of Canny edge detection is used, and then the resultant reflection and boundary mask free image is used to locate the approximate iris center using randomized Hough transform and then the polar transform is used to find ROI.??

Segmentation Procedure

The technique of extracting the target region out of an image by maintaining the quality of an image is called segmentation. The most crucial stage in this whole process of iris recognition is iris segmentation. In this paper, an algorithm of iris segmentation is presented which is designed based on the natural properties of the iris under general conditions. The first and the most important step in this algorithm is iris segmentation. The main objective of iris segmentation is to draw out the major distinguishing patterns in iris, for which we need to separate out noise and other minor patterns in iris. If we are able to find these major distinguishing patterns accurately, then the accuracy of uniquely recognizing using iris also increases exponentially.

Conclusion

In general, iris segmentation is built according to a particular database. But this leads to the lack of scalability. In this paper, we have proposed a two-stage iris segmentation framework. The segmentation algorithms are measured or compared based on the three main factors also known as quality factors, accuracy, usability, and speed. The proposed two-stage algorithm requires less processing time. It provides a robust method for the iris-based image segmentation. Upon experimenting, we found that the proposed method is scalable, and it does not depend on the iris database used. Hence, the proposed method of randomized Hough transform increases efficiency both in terms of time and space as compared to existing database-dependent iris segmentation procedures.

Continue reading from our below book chapter