Automated banknote recognition systems are essential for people with visual impairments who face challenges distinguishing between different currency denominations. This study presents a novel method aimed at helping blind people identify banknotes from three different countries (Egypt, Saudi Arabia, and the United States of America) by using a proposed feature detection algorithm. Our proposed system has two main stages: the proposed algorithm uses the Speeded-UP Robust Features (SURF) algorithm for key point detection, as it is fast and robust to variations in geometry and lighting. Then, it extracts features using the scale-invariant feature transform (SIFT) and histogram of oriented gradients (HOG) algorithms, which are scale invariant. This algorithm aims to overcome the limitations of both the SURF and SIFT algorithms and reduce the average response time and computational cost of the SIFT and HOG algorithms. We developed a banknote dataset with 12 classes for three countries. The accuracy of the banknote recognition was 99.2%. The performance of the proposed dataset was compared with that of the global Kaggle Egyptian dataset, resulting in 98.9% accuracy