The paper introduces SURF (Speeded Up Robust Features), a novel scale- and rotation-invariant interest point detector and descriptor. SURF aims to achieve faster computation and better performance compared to existing methods while maintaining or improving repeatability, distinctiveness, and robustness. The detector is based on the Hessian matrix, using integral images to reduce computation time, and the descriptor is a distribution of Haar-wavelet responses within the interest point neighborhood. The descriptor is simplified to 64 dimensions to balance speed and robustness. The paper presents experimental results on standard evaluation sets and a real-life object recognition application, demonstrating SURF's strong performance in both speed and accuracy. The SURF descriptor outperforms other state-of-the-art descriptors in recall-(1-precision) graphs and is significantly faster to compute. The paper also discusses the implementation details of the detector and descriptor, including the use of integral images and the orientation assignment process.The paper introduces SURF (Speeded Up Robust Features), a novel scale- and rotation-invariant interest point detector and descriptor. SURF aims to achieve faster computation and better performance compared to existing methods while maintaining or improving repeatability, distinctiveness, and robustness. The detector is based on the Hessian matrix, using integral images to reduce computation time, and the descriptor is a distribution of Haar-wavelet responses within the interest point neighborhood. The descriptor is simplified to 64 dimensions to balance speed and robustness. The paper presents experimental results on standard evaluation sets and a real-life object recognition application, demonstrating SURF's strong performance in both speed and accuracy. The SURF descriptor outperforms other state-of-the-art descriptors in recall-(1-precision) graphs and is significantly faster to compute. The paper also discusses the implementation details of the detector and descriptor, including the use of integral images and the orientation assignment process.