Fractal Triangular Search speeds up image content retrieval by 22%

A new research paper from Rodrigues et al. introduces Fractal Triangular Search (FTS), a metaheuristic designed to accelerate image content search. Unlike traditional methods that scan blindly, FTS frames content discovery as an optimization problem: it looks for 'evidence elements' that are spatially correlated with the target content. The local search algorithm follows a fractal pattern—a chain of triangles that engulf each other and grow indefinitely while shifting orientation each iteration. This variable neighborhood search explores the image more efficiently, requiring fewer visits to incorrect locations.

In extensive experiments, FTS consistently beat state-of-the-art metaheuristics. In the first group of nine test cases, FTS was faster in seven, averaging over 8% speed gain versus the runner-up. In the second group of seven cases, FTS led in six with an average improvement of over 22%. The advantage scales with image size—the larger the image, the more FTS outperforms competitors. This makes FTS promising for large-scale image databases and real-time computer vision tasks where speed and accuracy are critical.