Iris Detection by Discrete Sine Transform Based Feature Vector Using Random Forest
Shahid Akbar, Ashfaq Ahmad, Maqsood Hayat · 2014
The significance of Iris detection and recognition has been increased from last few decades. Looking at the importance of Iris detection and recognition, we propose a robust, stable and reliable computational model. Three Feature extraction strategies including Discrete Sine Transform ( DST ), Hilbert transform and Fast wavelet Hadamard Transform are used in order to extract numerical descriptors from iris images. Random forest is utilized as a learner. 5-folds cross validation test is applied to evaluate the performance of Random Forest. Among three feature spaces, DST feature space has achieved promising results. The success rate of Random forest on DST feature space is 93.4%. After examining the results, we have observed that our model might be useful and helpful for iris detection in future work.