Accurate Pupil and Iris Localization Using Least Mean Square Error Method

This paper presents a comprehensive iris recognition methodology utilizing the Least Mean Square (LMS) error minimization approach. The implementation involves multiple computational stages: initial pupil localization through adaptive thresholding and contour detection, outer iris boundary identification using circular Hough transform variants, and intensity normalization through Daugman's rubber sheet model. Image enhancement techniques including histogram equalization and contrast-limited adaptive histogram equalization (CLAHE) are applied to improve feature extraction reliability. The system employs iterative circle fitting algorithms that minimize the mean square error between detected boundaries and ideal circular models, significantly improving localization precision. Key MATLAB functions implemented include regionprops for blob analysis, imfindcircles for circular detection, and custom optimization routines for error minimization. This integrated approach achieves superior accuracy in biometric identification tasks through robust preprocessing and mathematical optimization techniques. Through systematic application of these algorithms, the methodology demonstrates significant improvements in iris recognition reliability, particularly in challenging imaging conditions with occlusions or non-ideal lighting. The complete pipeline ensures computational efficiency while maintaining high biometric authentication standards required for international security applications.