Bouncing and raindrop image search algorithms, two novel feature detection mechanisms

In this paper, two novel image search mechanisms are introduced and compared against the traditional linear approach and one another. These feature detection algorithms mimic two natural human search strategies which make them suitable for faster feature detection within pixelated images. Unlike linear approaches where pixel-by-pixel probing is required, these mechanisms provide faster feature detection due to their non-linear nature. The proposed algorithms are suitable for various feature detection applications, such as face detection, image labelling, abnormality detection in medical images, and the like. Raindrop and Bouncing mechanisms outperformed the linear approach by 75.2% and 89.8% respectively when tested separately against the linear algorithm. Moreover, after conducting extensive experiments with various conditions on all three algorithms, Raindrop detected 50.85% of the features while Bouncing and linear algorithms were successful in 35.08% and 14.07% of the tests respectively.