Precision Segmentation of Iris and Sclera: Mitigating Noise and Occlusions in Biometric Identification

Abstract

Biometric identification has become more and more popular as a safe and dependable way to authenticate, and the special and steady qualities of iris and sclera recognition systems make them essential. Noise and obstructions, such eyelashes and eyelids, can, however, make precise iris and sclera segmentation difficult. By means of a precision segmentation technique that improves the robustness of biometric identification systems, this work tackles these problems. Two primary goals are the emphasis of the suggested strategy: First, by reducing noise and eliminating superfluous areas, it seeks to obtain precise iris and sclera segmenting. To this purpose, a fine-tuned random forest algorithm is used, which is especially designed to differentiate between the iris, sclera, and occlusions. To guarantee robustness and generalization over various eye images, this algorithm is trained on a varied dataset. Second, a fusion model incorporating ocular characteristics from the sclera and iris is presented in this work. Through the combination of these characteristics, the model produces a more unique and secure biometric template. Key feature vectors are taken out of segmented regions, combined, and kept in a repository for later identification tasks. This integrated method takes use of the combined uniqueness of the iris and sclera to improve the accuracy of the biometric system and to add another level of security. The MMU Iris Dataset and the SBVPI Dataset, both publicly available, are used in this work. Extensive experiments showing the efficacy of the suggested approach show that the iris and sclera regions were segmented with an accuracy of 80% (MMU Dataset) and 90% (SBVPI Dataset) by the fine-tuned random forest algorithm. With such accuracy over conventional techniques, feature fusion and sophisticated segmentation techniques have great promise to improve biometric identification systems. Finally, a new and practical method for enhancing the security and dependability of iris and sclera-based biometric systems is presented in this work. The suggested approach opens up more precise and safe biometric identification solutions by tackling the problems of noise and occlusion through sophisticated segmentation and feature fusion